From: Roman Lebedev <lebedev.ri@gmail.com>
Date: Tue, 13 Dec 2022 01:23:34 +0000 (+0300)
Subject: [NFC][Codegen][X86] Add higher-VF interleaved load codegen tests that got lost
X-Git-Tag: upstream/17.0.6~24128
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=a13a2d105b11208cfe3b4c00e1cdefe0c0076df1;p=platform%2Fupstream%2Fllvm.git

[NFC][Codegen][X86] Add higher-VF interleaved load codegen tests that got lost
---

diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-2.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-2.ll
index 186cfe5..f3edc9d 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-2.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-2.ll
@@ -34,13 +34,10 @@ define void @load_i16_stride2_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nou
 ; AVX-NEXT:    vmovd %xmm0, (%rdx)
 ; AVX-NEXT:    retq
   %wide.vec = load <4 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <4 x i16> %wide.vec, <4 x i16> poison, <2 x i32> <i32 0, i32 2>
   %strided.vec1 = shufflevector <4 x i16> %wide.vec, <4 x i16> poison, <2 x i32> <i32 1, i32 3>
-
   store <2 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i16> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -76,13 +73,10 @@ define void @load_i16_stride2_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nou
 ; AVX512-NEXT:    vmovq %xmm1, (%rdx)
 ; AVX512-NEXT:    retq
   %wide.vec = load <8 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <8 x i16> %wide.vec, <8 x i16> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
   %strided.vec1 = shufflevector <8 x i16> %wide.vec, <8 x i16> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
-
   store <4 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i16> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -131,13 +125,10 @@ define void @load_i16_stride2_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nou
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <16 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %strided.vec1 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
-
   store <8 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i16> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -293,13 +284,10 @@ define void @load_i16_stride2_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) no
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <32 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %strided.vec1 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
-
   store <16 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i16> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -571,13 +559,10 @@ define void @load_i16_stride2_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) no
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <64 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46, i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62>
   %strided.vec1 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <32 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31, i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47, i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63>
-
   store <32 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i16> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -1099,13 +1084,10 @@ define void @load_i16_stride2_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) no
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <128 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <64 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46, i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62, i32 64, i32 66, i32 68, i32 70, i32 72, i32 74, i32 76, i32 78, i32 80, i32 82, i32 84, i32 86, i32 88, i32 90, i32 92, i32 94, i32 96, i32 98, i32 100, i32 102, i32 104, i32 106, i32 108, i32 110, i32 112, i32 114, i32 116, i32 118, i32 120, i32 122, i32 124, i32 126>
   %strided.vec1 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <64 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31, i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47, i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63, i32 65, i32 67, i32 69, i32 71, i32 73, i32 75, i32 77, i32 79, i32 81, i32 83, i32 85, i32 87, i32 89, i32 91, i32 93, i32 95, i32 97, i32 99, i32 101, i32 103, i32 105, i32 107, i32 109, i32 111, i32 113, i32 115, i32 117, i32 119, i32 121, i32 123, i32 125, i32 127>
-
   store <64 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i16> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
index 79b3508..67370c6 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
@@ -125,15 +125,12 @@ define void @load_i16_stride3_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vmovd %xmm0, (%rcx)
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <6 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <6 x i16> %wide.vec, <6 x i16> poison, <2 x i32> <i32 0, i32 3>
   %strided.vec1 = shufflevector <6 x i16> %wide.vec, <6 x i16> poison, <2 x i32> <i32 1, i32 4>
   %strided.vec2 = shufflevector <6 x i16> %wide.vec, <6 x i16> poison, <2 x i32> <i32 2, i32 5>
-
   store <2 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i16> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -297,15 +294,12 @@ define void @load_i16_stride3_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vmovq %xmm4, (%rcx)
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <12 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <12 x i16> %wide.vec, <12 x i16> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
   %strided.vec1 = shufflevector <12 x i16> %wide.vec, <12 x i16> poison, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
   %strided.vec2 = shufflevector <12 x i16> %wide.vec, <12 x i16> poison, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
-
   store <4 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i16> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -442,15 +436,12 @@ define void @load_i16_stride3_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <24 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <24 x i16> %wide.vec, <24 x i16> poison, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
   %strided.vec1 = shufflevector <24 x i16> %wide.vec, <24 x i16> poison, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
   %strided.vec2 = shufflevector <24 x i16> %wide.vec, <24 x i16> poison, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
-
   store <8 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i16> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -706,15 +697,12 @@ define void @load_i16_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <48 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <48 x i16> %wide.vec, <48 x i16> poison, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
   %strided.vec1 = shufflevector <48 x i16> %wide.vec, <48 x i16> poison, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
   %strided.vec2 = shufflevector <48 x i16> %wide.vec, <48 x i16> poison, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
-
   store <16 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i16> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -1243,15 +1231,12 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <96 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <96 x i16> %wide.vec, <96 x i16> poison, <32 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45, i32 48, i32 51, i32 54, i32 57, i32 60, i32 63, i32 66, i32 69, i32 72, i32 75, i32 78, i32 81, i32 84, i32 87, i32 90, i32 93>
   %strided.vec1 = shufflevector <96 x i16> %wide.vec, <96 x i16> poison, <32 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46, i32 49, i32 52, i32 55, i32 58, i32 61, i32 64, i32 67, i32 70, i32 73, i32 76, i32 79, i32 82, i32 85, i32 88, i32 91, i32 94>
   %strided.vec2 = shufflevector <96 x i16> %wide.vec, <96 x i16> poison, <32 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47, i32 50, i32 53, i32 56, i32 59, i32 62, i32 65, i32 68, i32 71, i32 74, i32 77, i32 80, i32 83, i32 86, i32 89, i32 92, i32 95>
-
   store <32 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i16> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -2370,15 +2355,12 @@ define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <192 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <64 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45, i32 48, i32 51, i32 54, i32 57, i32 60, i32 63, i32 66, i32 69, i32 72, i32 75, i32 78, i32 81, i32 84, i32 87, i32 90, i32 93, i32 96, i32 99, i32 102, i32 105, i32 108, i32 111, i32 114, i32 117, i32 120, i32 123, i32 126, i32 129, i32 132, i32 135, i32 138, i32 141, i32 144, i32 147, i32 150, i32 153, i32 156, i32 159, i32 162, i32 165, i32 168, i32 171, i32 174, i32 177, i32 180, i32 183, i32 186, i32 189>
   %strided.vec1 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <64 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46, i32 49, i32 52, i32 55, i32 58, i32 61, i32 64, i32 67, i32 70, i32 73, i32 76, i32 79, i32 82, i32 85, i32 88, i32 91, i32 94, i32 97, i32 100, i32 103, i32 106, i32 109, i32 112, i32 115, i32 118, i32 121, i32 124, i32 127, i32 130, i32 133, i32 136, i32 139, i32 142, i32 145, i32 148, i32 151, i32 154, i32 157, i32 160, i32 163, i32 166, i32 169, i32 172, i32 175, i32 178, i32 181, i32 184, i32 187, i32 190>
   %strided.vec2 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <64 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47, i32 50, i32 53, i32 56, i32 59, i32 62, i32 65, i32 68, i32 71, i32 74, i32 77, i32 80, i32 83, i32 86, i32 89, i32 92, i32 95, i32 98, i32 101, i32 104, i32 107, i32 110, i32 113, i32 116, i32 119, i32 122, i32 125, i32 128, i32 131, i32 134, i32 137, i32 140, i32 143, i32 146, i32 149, i32 152, i32 155, i32 158, i32 161, i32 164, i32 167, i32 170, i32 173, i32 176, i32 179, i32 182, i32 185, i32 188, i32 191>
-
   store <64 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i16> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll
index 3767d9d..5f692f4 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll
@@ -100,17 +100,14 @@ define void @load_i16_stride4_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vmovd %xmm2, (%r8)
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <8 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <8 x i16> %wide.vec, <8 x i16> poison, <2 x i32> <i32 0, i32 4>
   %strided.vec1 = shufflevector <8 x i16> %wide.vec, <8 x i16> poison, <2 x i32> <i32 1, i32 5>
   %strided.vec2 = shufflevector <8 x i16> %wide.vec, <8 x i16> poison, <2 x i32> <i32 2, i32 6>
   %strided.vec3 = shufflevector <8 x i16> %wide.vec, <8 x i16> poison, <2 x i32> <i32 3, i32 7>
-
   store <2 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <2 x i16> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -322,17 +319,14 @@ define void @load_i16_stride4_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <16 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
   %strided.vec1 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
   %strided.vec2 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
   %strided.vec3 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
-
   store <4 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <4 x i16> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -682,17 +676,14 @@ define void @load_i16_stride4_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <32 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
   %strided.vec1 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
   %strided.vec2 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
   %strided.vec3 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
-
   store <8 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <8 x i16> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -1357,17 +1348,14 @@ define void @load_i16_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <64 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
   %strided.vec1 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
   %strided.vec2 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
   %strided.vec3 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
-
   store <16 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <16 x i16> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -2842,17 +2830,14 @@ define void @load_i16_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <128 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <32 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124>
   %strided.vec1 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <32 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125>
   %strided.vec2 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <32 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126>
   %strided.vec3 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <32 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127>
-
   store <32 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <32 x i16> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -6081,17 +6066,14 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <256 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <64 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124, i32 128, i32 132, i32 136, i32 140, i32 144, i32 148, i32 152, i32 156, i32 160, i32 164, i32 168, i32 172, i32 176, i32 180, i32 184, i32 188, i32 192, i32 196, i32 200, i32 204, i32 208, i32 212, i32 216, i32 220, i32 224, i32 228, i32 232, i32 236, i32 240, i32 244, i32 248, i32 252>
   %strided.vec1 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <64 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125, i32 129, i32 133, i32 137, i32 141, i32 145, i32 149, i32 153, i32 157, i32 161, i32 165, i32 169, i32 173, i32 177, i32 181, i32 185, i32 189, i32 193, i32 197, i32 201, i32 205, i32 209, i32 213, i32 217, i32 221, i32 225, i32 229, i32 233, i32 237, i32 241, i32 245, i32 249, i32 253>
   %strided.vec2 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <64 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126, i32 130, i32 134, i32 138, i32 142, i32 146, i32 150, i32 154, i32 158, i32 162, i32 166, i32 170, i32 174, i32 178, i32 182, i32 186, i32 190, i32 194, i32 198, i32 202, i32 206, i32 210, i32 214, i32 218, i32 222, i32 226, i32 230, i32 234, i32 238, i32 242, i32 246, i32 250, i32 254>
   %strided.vec3 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <64 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127, i32 131, i32 135, i32 139, i32 143, i32 147, i32 151, i32 155, i32 159, i32 163, i32 167, i32 171, i32 175, i32 179, i32 183, i32 187, i32 191, i32 195, i32 199, i32 203, i32 207, i32 211, i32 215, i32 219, i32 223, i32 227, i32 231, i32 235, i32 239, i32 243, i32 247, i32 251, i32 255>
-
   store <64 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <64 x i16> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
index 6cd7ae7..de2ff07 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
@@ -194,19 +194,16 @@ define void @load_i16_stride5_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vmovd %xmm1, (%r9)
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <10 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <10 x i16> %wide.vec, <10 x i16> poison, <2 x i32> <i32 0, i32 5>
   %strided.vec1 = shufflevector <10 x i16> %wide.vec, <10 x i16> poison, <2 x i32> <i32 1, i32 6>
   %strided.vec2 = shufflevector <10 x i16> %wide.vec, <10 x i16> poison, <2 x i32> <i32 2, i32 7>
   %strided.vec3 = shufflevector <10 x i16> %wide.vec, <10 x i16> poison, <2 x i32> <i32 3, i32 8>
   %strided.vec4 = shufflevector <10 x i16> %wide.vec, <10 x i16> poison, <2 x i32> <i32 4, i32 9>
-
   store <2 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <2 x i16> %strided.vec3, ptr %out.vec3, align 64
   store <2 x i16> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -476,19 +473,16 @@ define void @load_i16_stride5_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <20 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <20 x i16> %wide.vec, <20 x i16> poison, <4 x i32> <i32 0, i32 5, i32 10, i32 15>
   %strided.vec1 = shufflevector <20 x i16> %wide.vec, <20 x i16> poison, <4 x i32> <i32 1, i32 6, i32 11, i32 16>
   %strided.vec2 = shufflevector <20 x i16> %wide.vec, <20 x i16> poison, <4 x i32> <i32 2, i32 7, i32 12, i32 17>
   %strided.vec3 = shufflevector <20 x i16> %wide.vec, <20 x i16> poison, <4 x i32> <i32 3, i32 8, i32 13, i32 18>
   %strided.vec4 = shufflevector <20 x i16> %wide.vec, <20 x i16> poison, <4 x i32> <i32 4, i32 9, i32 14, i32 19>
-
   store <4 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <4 x i16> %strided.vec3, ptr %out.vec3, align 64
   store <4 x i16> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -1093,19 +1087,16 @@ define void @load_i16_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <40 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <40 x i16> %wide.vec, <40 x i16> poison, <8 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35>
   %strided.vec1 = shufflevector <40 x i16> %wide.vec, <40 x i16> poison, <8 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36>
   %strided.vec2 = shufflevector <40 x i16> %wide.vec, <40 x i16> poison, <8 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37>
   %strided.vec3 = shufflevector <40 x i16> %wide.vec, <40 x i16> poison, <8 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38>
   %strided.vec4 = shufflevector <40 x i16> %wide.vec, <40 x i16> poison, <8 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39>
-
   store <8 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <8 x i16> %strided.vec3, ptr %out.vec3, align 64
   store <8 x i16> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -1939,19 +1930,16 @@ define void @load_i16_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <80 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <80 x i16> %wide.vec, <80 x i16> poison, <16 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75>
   %strided.vec1 = shufflevector <80 x i16> %wide.vec, <80 x i16> poison, <16 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76>
   %strided.vec2 = shufflevector <80 x i16> %wide.vec, <80 x i16> poison, <16 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77>
   %strided.vec3 = shufflevector <80 x i16> %wide.vec, <80 x i16> poison, <16 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78>
   %strided.vec4 = shufflevector <80 x i16> %wide.vec, <80 x i16> poison, <16 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79>
-
   store <16 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <16 x i16> %strided.vec3, ptr %out.vec3, align 64
   store <16 x i16> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -3839,19 +3827,16 @@ define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <160 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75, i32 80, i32 85, i32 90, i32 95, i32 100, i32 105, i32 110, i32 115, i32 120, i32 125, i32 130, i32 135, i32 140, i32 145, i32 150, i32 155>
   %strided.vec1 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76, i32 81, i32 86, i32 91, i32 96, i32 101, i32 106, i32 111, i32 116, i32 121, i32 126, i32 131, i32 136, i32 141, i32 146, i32 151, i32 156>
   %strided.vec2 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77, i32 82, i32 87, i32 92, i32 97, i32 102, i32 107, i32 112, i32 117, i32 122, i32 127, i32 132, i32 137, i32 142, i32 147, i32 152, i32 157>
   %strided.vec3 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78, i32 83, i32 88, i32 93, i32 98, i32 103, i32 108, i32 113, i32 118, i32 123, i32 128, i32 133, i32 138, i32 143, i32 148, i32 153, i32 158>
   %strided.vec4 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79, i32 84, i32 89, i32 94, i32 99, i32 104, i32 109, i32 114, i32 119, i32 124, i32 129, i32 134, i32 139, i32 144, i32 149, i32 154, i32 159>
-
   store <32 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <32 x i16> %strided.vec3, ptr %out.vec3, align 64
   store <32 x i16> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -7936,19 +7921,16 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <320 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <320 x i16> %wide.vec, <320 x i16> poison, <64 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75, i32 80, i32 85, i32 90, i32 95, i32 100, i32 105, i32 110, i32 115, i32 120, i32 125, i32 130, i32 135, i32 140, i32 145, i32 150, i32 155, i32 160, i32 165, i32 170, i32 175, i32 180, i32 185, i32 190, i32 195, i32 200, i32 205, i32 210, i32 215, i32 220, i32 225, i32 230, i32 235, i32 240, i32 245, i32 250, i32 255, i32 260, i32 265, i32 270, i32 275, i32 280, i32 285, i32 290, i32 295, i32 300, i32 305, i32 310, i32 315>
   %strided.vec1 = shufflevector <320 x i16> %wide.vec, <320 x i16> poison, <64 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76, i32 81, i32 86, i32 91, i32 96, i32 101, i32 106, i32 111, i32 116, i32 121, i32 126, i32 131, i32 136, i32 141, i32 146, i32 151, i32 156, i32 161, i32 166, i32 171, i32 176, i32 181, i32 186, i32 191, i32 196, i32 201, i32 206, i32 211, i32 216, i32 221, i32 226, i32 231, i32 236, i32 241, i32 246, i32 251, i32 256, i32 261, i32 266, i32 271, i32 276, i32 281, i32 286, i32 291, i32 296, i32 301, i32 306, i32 311, i32 316>
   %strided.vec2 = shufflevector <320 x i16> %wide.vec, <320 x i16> poison, <64 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77, i32 82, i32 87, i32 92, i32 97, i32 102, i32 107, i32 112, i32 117, i32 122, i32 127, i32 132, i32 137, i32 142, i32 147, i32 152, i32 157, i32 162, i32 167, i32 172, i32 177, i32 182, i32 187, i32 192, i32 197, i32 202, i32 207, i32 212, i32 217, i32 222, i32 227, i32 232, i32 237, i32 242, i32 247, i32 252, i32 257, i32 262, i32 267, i32 272, i32 277, i32 282, i32 287, i32 292, i32 297, i32 302, i32 307, i32 312, i32 317>
   %strided.vec3 = shufflevector <320 x i16> %wide.vec, <320 x i16> poison, <64 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78, i32 83, i32 88, i32 93, i32 98, i32 103, i32 108, i32 113, i32 118, i32 123, i32 128, i32 133, i32 138, i32 143, i32 148, i32 153, i32 158, i32 163, i32 168, i32 173, i32 178, i32 183, i32 188, i32 193, i32 198, i32 203, i32 208, i32 213, i32 218, i32 223, i32 228, i32 233, i32 238, i32 243, i32 248, i32 253, i32 258, i32 263, i32 268, i32 273, i32 278, i32 283, i32 288, i32 293, i32 298, i32 303, i32 308, i32 313, i32 318>
   %strided.vec4 = shufflevector <320 x i16> %wide.vec, <320 x i16> poison, <64 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79, i32 84, i32 89, i32 94, i32 99, i32 104, i32 109, i32 114, i32 119, i32 124, i32 129, i32 134, i32 139, i32 144, i32 149, i32 154, i32 159, i32 164, i32 169, i32 174, i32 179, i32 184, i32 189, i32 194, i32 199, i32 204, i32 209, i32 214, i32 219, i32 224, i32 229, i32 234, i32 239, i32 244, i32 249, i32 254, i32 259, i32 264, i32 269, i32 274, i32 279, i32 284, i32 289, i32 294, i32 299, i32 304, i32 309, i32 314, i32 319>
-
   store <64 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <64 x i16> %strided.vec3, ptr %out.vec3, align 64
   store <64 x i16> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
index c2261a5..b7465fb 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
@@ -256,21 +256,18 @@ define void @load_i16_stride6_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vmovd %xmm0, (%rax)
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <12 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <12 x i16> %wide.vec, <12 x i16> poison, <2 x i32> <i32 0, i32 6>
   %strided.vec1 = shufflevector <12 x i16> %wide.vec, <12 x i16> poison, <2 x i32> <i32 1, i32 7>
   %strided.vec2 = shufflevector <12 x i16> %wide.vec, <12 x i16> poison, <2 x i32> <i32 2, i32 8>
   %strided.vec3 = shufflevector <12 x i16> %wide.vec, <12 x i16> poison, <2 x i32> <i32 3, i32 9>
   %strided.vec4 = shufflevector <12 x i16> %wide.vec, <12 x i16> poison, <2 x i32> <i32 4, i32 10>
   %strided.vec5 = shufflevector <12 x i16> %wide.vec, <12 x i16> poison, <2 x i32> <i32 5, i32 11>
-
   store <2 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <2 x i16> %strided.vec3, ptr %out.vec3, align 64
   store <2 x i16> %strided.vec4, ptr %out.vec4, align 64
   store <2 x i16> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -554,21 +551,18 @@ define void @load_i16_stride6_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <24 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <24 x i16> %wide.vec, <24 x i16> poison, <4 x i32> <i32 0, i32 6, i32 12, i32 18>
   %strided.vec1 = shufflevector <24 x i16> %wide.vec, <24 x i16> poison, <4 x i32> <i32 1, i32 7, i32 13, i32 19>
   %strided.vec2 = shufflevector <24 x i16> %wide.vec, <24 x i16> poison, <4 x i32> <i32 2, i32 8, i32 14, i32 20>
   %strided.vec3 = shufflevector <24 x i16> %wide.vec, <24 x i16> poison, <4 x i32> <i32 3, i32 9, i32 15, i32 21>
   %strided.vec4 = shufflevector <24 x i16> %wide.vec, <24 x i16> poison, <4 x i32> <i32 4, i32 10, i32 16, i32 22>
   %strided.vec5 = shufflevector <24 x i16> %wide.vec, <24 x i16> poison, <4 x i32> <i32 5, i32 11, i32 17, i32 23>
-
   store <4 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <4 x i16> %strided.vec3, ptr %out.vec3, align 64
   store <4 x i16> %strided.vec4, ptr %out.vec4, align 64
   store <4 x i16> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -1344,21 +1338,18 @@ define void @load_i16_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <48 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <48 x i16> %wide.vec, <48 x i16> poison, <8 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42>
   %strided.vec1 = shufflevector <48 x i16> %wide.vec, <48 x i16> poison, <8 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43>
   %strided.vec2 = shufflevector <48 x i16> %wide.vec, <48 x i16> poison, <8 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44>
   %strided.vec3 = shufflevector <48 x i16> %wide.vec, <48 x i16> poison, <8 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45>
   %strided.vec4 = shufflevector <48 x i16> %wide.vec, <48 x i16> poison, <8 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46>
   %strided.vec5 = shufflevector <48 x i16> %wide.vec, <48 x i16> poison, <8 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47>
-
   store <8 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <8 x i16> %strided.vec3, ptr %out.vec3, align 64
   store <8 x i16> %strided.vec4, ptr %out.vec4, align 64
   store <8 x i16> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -2536,21 +2527,18 @@ define void @load_i16_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <96 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <96 x i16> %wide.vec, <96 x i16> poison, <16 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90>
   %strided.vec1 = shufflevector <96 x i16> %wide.vec, <96 x i16> poison, <16 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91>
   %strided.vec2 = shufflevector <96 x i16> %wide.vec, <96 x i16> poison, <16 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92>
   %strided.vec3 = shufflevector <96 x i16> %wide.vec, <96 x i16> poison, <16 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93>
   %strided.vec4 = shufflevector <96 x i16> %wide.vec, <96 x i16> poison, <16 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94>
   %strided.vec5 = shufflevector <96 x i16> %wide.vec, <96 x i16> poison, <16 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95>
-
   store <16 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <16 x i16> %strided.vec3, ptr %out.vec3, align 64
   store <16 x i16> %strided.vec4, ptr %out.vec4, align 64
   store <16 x i16> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -5903,21 +5891,18 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <192 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <32 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90, i32 96, i32 102, i32 108, i32 114, i32 120, i32 126, i32 132, i32 138, i32 144, i32 150, i32 156, i32 162, i32 168, i32 174, i32 180, i32 186>
   %strided.vec1 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <32 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91, i32 97, i32 103, i32 109, i32 115, i32 121, i32 127, i32 133, i32 139, i32 145, i32 151, i32 157, i32 163, i32 169, i32 175, i32 181, i32 187>
   %strided.vec2 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <32 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92, i32 98, i32 104, i32 110, i32 116, i32 122, i32 128, i32 134, i32 140, i32 146, i32 152, i32 158, i32 164, i32 170, i32 176, i32 182, i32 188>
   %strided.vec3 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <32 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93, i32 99, i32 105, i32 111, i32 117, i32 123, i32 129, i32 135, i32 141, i32 147, i32 153, i32 159, i32 165, i32 171, i32 177, i32 183, i32 189>
   %strided.vec4 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <32 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94, i32 100, i32 106, i32 112, i32 118, i32 124, i32 130, i32 136, i32 142, i32 148, i32 154, i32 160, i32 166, i32 172, i32 178, i32 184, i32 190>
   %strided.vec5 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <32 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95, i32 101, i32 107, i32 113, i32 119, i32 125, i32 131, i32 137, i32 143, i32 149, i32 155, i32 161, i32 167, i32 173, i32 179, i32 185, i32 191>
-
   store <32 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <32 x i16> %strided.vec3, ptr %out.vec3, align 64
   store <32 x i16> %strided.vec4, ptr %out.vec4, align 64
   store <32 x i16> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -13043,21 +13028,18 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <384 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <384 x i16> %wide.vec, <384 x i16> poison, <64 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90, i32 96, i32 102, i32 108, i32 114, i32 120, i32 126, i32 132, i32 138, i32 144, i32 150, i32 156, i32 162, i32 168, i32 174, i32 180, i32 186, i32 192, i32 198, i32 204, i32 210, i32 216, i32 222, i32 228, i32 234, i32 240, i32 246, i32 252, i32 258, i32 264, i32 270, i32 276, i32 282, i32 288, i32 294, i32 300, i32 306, i32 312, i32 318, i32 324, i32 330, i32 336, i32 342, i32 348, i32 354, i32 360, i32 366, i32 372, i32 378>
   %strided.vec1 = shufflevector <384 x i16> %wide.vec, <384 x i16> poison, <64 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91, i32 97, i32 103, i32 109, i32 115, i32 121, i32 127, i32 133, i32 139, i32 145, i32 151, i32 157, i32 163, i32 169, i32 175, i32 181, i32 187, i32 193, i32 199, i32 205, i32 211, i32 217, i32 223, i32 229, i32 235, i32 241, i32 247, i32 253, i32 259, i32 265, i32 271, i32 277, i32 283, i32 289, i32 295, i32 301, i32 307, i32 313, i32 319, i32 325, i32 331, i32 337, i32 343, i32 349, i32 355, i32 361, i32 367, i32 373, i32 379>
   %strided.vec2 = shufflevector <384 x i16> %wide.vec, <384 x i16> poison, <64 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92, i32 98, i32 104, i32 110, i32 116, i32 122, i32 128, i32 134, i32 140, i32 146, i32 152, i32 158, i32 164, i32 170, i32 176, i32 182, i32 188, i32 194, i32 200, i32 206, i32 212, i32 218, i32 224, i32 230, i32 236, i32 242, i32 248, i32 254, i32 260, i32 266, i32 272, i32 278, i32 284, i32 290, i32 296, i32 302, i32 308, i32 314, i32 320, i32 326, i32 332, i32 338, i32 344, i32 350, i32 356, i32 362, i32 368, i32 374, i32 380>
   %strided.vec3 = shufflevector <384 x i16> %wide.vec, <384 x i16> poison, <64 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93, i32 99, i32 105, i32 111, i32 117, i32 123, i32 129, i32 135, i32 141, i32 147, i32 153, i32 159, i32 165, i32 171, i32 177, i32 183, i32 189, i32 195, i32 201, i32 207, i32 213, i32 219, i32 225, i32 231, i32 237, i32 243, i32 249, i32 255, i32 261, i32 267, i32 273, i32 279, i32 285, i32 291, i32 297, i32 303, i32 309, i32 315, i32 321, i32 327, i32 333, i32 339, i32 345, i32 351, i32 357, i32 363, i32 369, i32 375, i32 381>
   %strided.vec4 = shufflevector <384 x i16> %wide.vec, <384 x i16> poison, <64 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94, i32 100, i32 106, i32 112, i32 118, i32 124, i32 130, i32 136, i32 142, i32 148, i32 154, i32 160, i32 166, i32 172, i32 178, i32 184, i32 190, i32 196, i32 202, i32 208, i32 214, i32 220, i32 226, i32 232, i32 238, i32 244, i32 250, i32 256, i32 262, i32 268, i32 274, i32 280, i32 286, i32 292, i32 298, i32 304, i32 310, i32 316, i32 322, i32 328, i32 334, i32 340, i32 346, i32 352, i32 358, i32 364, i32 370, i32 376, i32 382>
   %strided.vec5 = shufflevector <384 x i16> %wide.vec, <384 x i16> poison, <64 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95, i32 101, i32 107, i32 113, i32 119, i32 125, i32 131, i32 137, i32 143, i32 149, i32 155, i32 161, i32 167, i32 173, i32 179, i32 185, i32 191, i32 197, i32 203, i32 209, i32 215, i32 221, i32 227, i32 233, i32 239, i32 245, i32 251, i32 257, i32 263, i32 269, i32 275, i32 281, i32 287, i32 293, i32 299, i32 305, i32 311, i32 317, i32 323, i32 329, i32 335, i32 341, i32 347, i32 353, i32 359, i32 365, i32 371, i32 377, i32 383>
-
   store <64 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i16> %strided.vec2, ptr %out.vec2, align 64
   store <64 x i16> %strided.vec3, ptr %out.vec3, align 64
   store <64 x i16> %strided.vec4, ptr %out.vec4, align 64
   store <64 x i16> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
index 228c672..7b78444 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
@@ -281,7 +281,6 @@ define void @load_i16_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vmovd %xmm8, (%rax)
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <14 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <14 x i16> %wide.vec, <14 x i16> poison, <2 x i32> <i32 0, i32 7>
   %strided.vec1 = shufflevector <14 x i16> %wide.vec, <14 x i16> poison, <2 x i32> <i32 1, i32 8>
   %strided.vec2 = shufflevector <14 x i16> %wide.vec, <14 x i16> poison, <2 x i32> <i32 2, i32 9>
@@ -289,7 +288,6 @@ define void @load_i16_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <14 x i16> %wide.vec, <14 x i16> poison, <2 x i32> <i32 4, i32 11>
   %strided.vec5 = shufflevector <14 x i16> %wide.vec, <14 x i16> poison, <2 x i32> <i32 5, i32 12>
   %strided.vec6 = shufflevector <14 x i16> %wide.vec, <14 x i16> poison, <2 x i32> <i32 6, i32 13>
-
   store <2 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i16> %strided.vec2, ptr %out.vec2, align 64
@@ -297,7 +295,6 @@ define void @load_i16_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <2 x i16> %strided.vec4, ptr %out.vec4, align 64
   store <2 x i16> %strided.vec5, ptr %out.vec5, align 64
   store <2 x i16> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -717,7 +714,6 @@ define void @load_i16_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <28 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <28 x i16> %wide.vec, <28 x i16> poison, <4 x i32> <i32 0, i32 7, i32 14, i32 21>
   %strided.vec1 = shufflevector <28 x i16> %wide.vec, <28 x i16> poison, <4 x i32> <i32 1, i32 8, i32 15, i32 22>
   %strided.vec2 = shufflevector <28 x i16> %wide.vec, <28 x i16> poison, <4 x i32> <i32 2, i32 9, i32 16, i32 23>
@@ -725,7 +721,6 @@ define void @load_i16_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <28 x i16> %wide.vec, <28 x i16> poison, <4 x i32> <i32 4, i32 11, i32 18, i32 25>
   %strided.vec5 = shufflevector <28 x i16> %wide.vec, <28 x i16> poison, <4 x i32> <i32 5, i32 12, i32 19, i32 26>
   %strided.vec6 = shufflevector <28 x i16> %wide.vec, <28 x i16> poison, <4 x i32> <i32 6, i32 13, i32 20, i32 27>
-
   store <4 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i16> %strided.vec2, ptr %out.vec2, align 64
@@ -733,7 +728,6 @@ define void @load_i16_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <4 x i16> %strided.vec4, ptr %out.vec4, align 64
   store <4 x i16> %strided.vec5, ptr %out.vec5, align 64
   store <4 x i16> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -1743,7 +1737,6 @@ define void @load_i16_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <56 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <56 x i16> %wide.vec, <56 x i16> poison, <8 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49>
   %strided.vec1 = shufflevector <56 x i16> %wide.vec, <56 x i16> poison, <8 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50>
   %strided.vec2 = shufflevector <56 x i16> %wide.vec, <56 x i16> poison, <8 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51>
@@ -1751,7 +1744,6 @@ define void @load_i16_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <56 x i16> %wide.vec, <56 x i16> poison, <8 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53>
   %strided.vec5 = shufflevector <56 x i16> %wide.vec, <56 x i16> poison, <8 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54>
   %strided.vec6 = shufflevector <56 x i16> %wide.vec, <56 x i16> poison, <8 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55>
-
   store <8 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i16> %strided.vec2, ptr %out.vec2, align 64
@@ -1759,7 +1751,6 @@ define void @load_i16_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <8 x i16> %strided.vec4, ptr %out.vec4, align 64
   store <8 x i16> %strided.vec5, ptr %out.vec5, align 64
   store <8 x i16> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -3339,7 +3330,6 @@ define void @load_i16_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <112 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <112 x i16> %wide.vec, <112 x i16> poison, <16 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49, i32 56, i32 63, i32 70, i32 77, i32 84, i32 91, i32 98, i32 105>
   %strided.vec1 = shufflevector <112 x i16> %wide.vec, <112 x i16> poison, <16 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50, i32 57, i32 64, i32 71, i32 78, i32 85, i32 92, i32 99, i32 106>
   %strided.vec2 = shufflevector <112 x i16> %wide.vec, <112 x i16> poison, <16 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51, i32 58, i32 65, i32 72, i32 79, i32 86, i32 93, i32 100, i32 107>
@@ -3347,7 +3337,6 @@ define void @load_i16_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   %strided.vec4 = shufflevector <112 x i16> %wide.vec, <112 x i16> poison, <16 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53, i32 60, i32 67, i32 74, i32 81, i32 88, i32 95, i32 102, i32 109>
   %strided.vec5 = shufflevector <112 x i16> %wide.vec, <112 x i16> poison, <16 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54, i32 61, i32 68, i32 75, i32 82, i32 89, i32 96, i32 103, i32 110>
   %strided.vec6 = shufflevector <112 x i16> %wide.vec, <112 x i16> poison, <16 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55, i32 62, i32 69, i32 76, i32 83, i32 90, i32 97, i32 104, i32 111>
-
   store <16 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i16> %strided.vec2, ptr %out.vec2, align 64
@@ -3355,7 +3344,6 @@ define void @load_i16_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   store <16 x i16> %strided.vec4, ptr %out.vec4, align 64
   store <16 x i16> %strided.vec5, ptr %out.vec5, align 64
   store <16 x i16> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -7479,7 +7467,6 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <224 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <224 x i16> %wide.vec, <224 x i16> poison, <32 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49, i32 56, i32 63, i32 70, i32 77, i32 84, i32 91, i32 98, i32 105, i32 112, i32 119, i32 126, i32 133, i32 140, i32 147, i32 154, i32 161, i32 168, i32 175, i32 182, i32 189, i32 196, i32 203, i32 210, i32 217>
   %strided.vec1 = shufflevector <224 x i16> %wide.vec, <224 x i16> poison, <32 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50, i32 57, i32 64, i32 71, i32 78, i32 85, i32 92, i32 99, i32 106, i32 113, i32 120, i32 127, i32 134, i32 141, i32 148, i32 155, i32 162, i32 169, i32 176, i32 183, i32 190, i32 197, i32 204, i32 211, i32 218>
   %strided.vec2 = shufflevector <224 x i16> %wide.vec, <224 x i16> poison, <32 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51, i32 58, i32 65, i32 72, i32 79, i32 86, i32 93, i32 100, i32 107, i32 114, i32 121, i32 128, i32 135, i32 142, i32 149, i32 156, i32 163, i32 170, i32 177, i32 184, i32 191, i32 198, i32 205, i32 212, i32 219>
@@ -7487,7 +7474,6 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   %strided.vec4 = shufflevector <224 x i16> %wide.vec, <224 x i16> poison, <32 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53, i32 60, i32 67, i32 74, i32 81, i32 88, i32 95, i32 102, i32 109, i32 116, i32 123, i32 130, i32 137, i32 144, i32 151, i32 158, i32 165, i32 172, i32 179, i32 186, i32 193, i32 200, i32 207, i32 214, i32 221>
   %strided.vec5 = shufflevector <224 x i16> %wide.vec, <224 x i16> poison, <32 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54, i32 61, i32 68, i32 75, i32 82, i32 89, i32 96, i32 103, i32 110, i32 117, i32 124, i32 131, i32 138, i32 145, i32 152, i32 159, i32 166, i32 173, i32 180, i32 187, i32 194, i32 201, i32 208, i32 215, i32 222>
   %strided.vec6 = shufflevector <224 x i16> %wide.vec, <224 x i16> poison, <32 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55, i32 62, i32 69, i32 76, i32 83, i32 90, i32 97, i32 104, i32 111, i32 118, i32 125, i32 132, i32 139, i32 146, i32 153, i32 160, i32 167, i32 174, i32 181, i32 188, i32 195, i32 202, i32 209, i32 216, i32 223>
-
   store <32 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i16> %strided.vec2, ptr %out.vec2, align 64
@@ -7495,7 +7481,6 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   store <32 x i16> %strided.vec4, ptr %out.vec4, align 64
   store <32 x i16> %strided.vec5, ptr %out.vec5, align 64
   store <32 x i16> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -16305,7 +16290,6 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <448 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <448 x i16> %wide.vec, <448 x i16> poison, <64 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49, i32 56, i32 63, i32 70, i32 77, i32 84, i32 91, i32 98, i32 105, i32 112, i32 119, i32 126, i32 133, i32 140, i32 147, i32 154, i32 161, i32 168, i32 175, i32 182, i32 189, i32 196, i32 203, i32 210, i32 217, i32 224, i32 231, i32 238, i32 245, i32 252, i32 259, i32 266, i32 273, i32 280, i32 287, i32 294, i32 301, i32 308, i32 315, i32 322, i32 329, i32 336, i32 343, i32 350, i32 357, i32 364, i32 371, i32 378, i32 385, i32 392, i32 399, i32 406, i32 413, i32 420, i32 427, i32 434, i32 441>
   %strided.vec1 = shufflevector <448 x i16> %wide.vec, <448 x i16> poison, <64 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50, i32 57, i32 64, i32 71, i32 78, i32 85, i32 92, i32 99, i32 106, i32 113, i32 120, i32 127, i32 134, i32 141, i32 148, i32 155, i32 162, i32 169, i32 176, i32 183, i32 190, i32 197, i32 204, i32 211, i32 218, i32 225, i32 232, i32 239, i32 246, i32 253, i32 260, i32 267, i32 274, i32 281, i32 288, i32 295, i32 302, i32 309, i32 316, i32 323, i32 330, i32 337, i32 344, i32 351, i32 358, i32 365, i32 372, i32 379, i32 386, i32 393, i32 400, i32 407, i32 414, i32 421, i32 428, i32 435, i32 442>
   %strided.vec2 = shufflevector <448 x i16> %wide.vec, <448 x i16> poison, <64 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51, i32 58, i32 65, i32 72, i32 79, i32 86, i32 93, i32 100, i32 107, i32 114, i32 121, i32 128, i32 135, i32 142, i32 149, i32 156, i32 163, i32 170, i32 177, i32 184, i32 191, i32 198, i32 205, i32 212, i32 219, i32 226, i32 233, i32 240, i32 247, i32 254, i32 261, i32 268, i32 275, i32 282, i32 289, i32 296, i32 303, i32 310, i32 317, i32 324, i32 331, i32 338, i32 345, i32 352, i32 359, i32 366, i32 373, i32 380, i32 387, i32 394, i32 401, i32 408, i32 415, i32 422, i32 429, i32 436, i32 443>
@@ -16313,7 +16297,6 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   %strided.vec4 = shufflevector <448 x i16> %wide.vec, <448 x i16> poison, <64 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53, i32 60, i32 67, i32 74, i32 81, i32 88, i32 95, i32 102, i32 109, i32 116, i32 123, i32 130, i32 137, i32 144, i32 151, i32 158, i32 165, i32 172, i32 179, i32 186, i32 193, i32 200, i32 207, i32 214, i32 221, i32 228, i32 235, i32 242, i32 249, i32 256, i32 263, i32 270, i32 277, i32 284, i32 291, i32 298, i32 305, i32 312, i32 319, i32 326, i32 333, i32 340, i32 347, i32 354, i32 361, i32 368, i32 375, i32 382, i32 389, i32 396, i32 403, i32 410, i32 417, i32 424, i32 431, i32 438, i32 445>
   %strided.vec5 = shufflevector <448 x i16> %wide.vec, <448 x i16> poison, <64 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54, i32 61, i32 68, i32 75, i32 82, i32 89, i32 96, i32 103, i32 110, i32 117, i32 124, i32 131, i32 138, i32 145, i32 152, i32 159, i32 166, i32 173, i32 180, i32 187, i32 194, i32 201, i32 208, i32 215, i32 222, i32 229, i32 236, i32 243, i32 250, i32 257, i32 264, i32 271, i32 278, i32 285, i32 292, i32 299, i32 306, i32 313, i32 320, i32 327, i32 334, i32 341, i32 348, i32 355, i32 362, i32 369, i32 376, i32 383, i32 390, i32 397, i32 404, i32 411, i32 418, i32 425, i32 432, i32 439, i32 446>
   %strided.vec6 = shufflevector <448 x i16> %wide.vec, <448 x i16> poison, <64 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55, i32 62, i32 69, i32 76, i32 83, i32 90, i32 97, i32 104, i32 111, i32 118, i32 125, i32 132, i32 139, i32 146, i32 153, i32 160, i32 167, i32 174, i32 181, i32 188, i32 195, i32 202, i32 209, i32 216, i32 223, i32 230, i32 237, i32 244, i32 251, i32 258, i32 265, i32 272, i32 279, i32 286, i32 293, i32 300, i32 307, i32 314, i32 321, i32 328, i32 335, i32 342, i32 349, i32 356, i32 363, i32 370, i32 377, i32 384, i32 391, i32 398, i32 405, i32 412, i32 419, i32 426, i32 433, i32 440, i32 447>
-
   store <64 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i16> %strided.vec2, ptr %out.vec2, align 64
@@ -16321,7 +16304,6 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   store <64 x i16> %strided.vec4, ptr %out.vec4, align 64
   store <64 x i16> %strided.vec5, ptr %out.vec5, align 64
   store <64 x i16> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll
index 316544c..6c8d41e 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll
@@ -61,7 +61,6 @@ define void @load_i16_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX-NEXT:    vpextrd $3, %xmm0, (%rax)
 ; AVX-NEXT:    retq
   %wide.vec = load <16 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 0, i32 8>
   %strided.vec1 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 1, i32 9>
   %strided.vec2 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 2, i32 10>
@@ -70,7 +69,6 @@ define void @load_i16_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 5, i32 13>
   %strided.vec6 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 6, i32 14>
   %strided.vec7 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 7, i32 15>
-
   store <2 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i16> %strided.vec2, ptr %out.vec2, align 64
@@ -79,7 +77,6 @@ define void @load_i16_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <2 x i16> %strided.vec5, ptr %out.vec5, align 64
   store <2 x i16> %strided.vec6, ptr %out.vec6, align 64
   store <2 x i16> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -303,7 +300,6 @@ define void @load_i16_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <32 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
   %strided.vec1 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 1, i32 9, i32 17, i32 25>
   %strided.vec2 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 2, i32 10, i32 18, i32 26>
@@ -312,7 +308,6 @@ define void @load_i16_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 5, i32 13, i32 21, i32 29>
   %strided.vec6 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 6, i32 14, i32 22, i32 30>
   %strided.vec7 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 7, i32 15, i32 23, i32 31>
-
   store <4 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i16> %strided.vec2, ptr %out.vec2, align 64
@@ -321,7 +316,6 @@ define void @load_i16_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <4 x i16> %strided.vec5, ptr %out.vec5, align 64
   store <4 x i16> %strided.vec6, ptr %out.vec6, align 64
   store <4 x i16> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -950,7 +944,6 @@ define void @load_i16_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <64 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56>
   %strided.vec1 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57>
   %strided.vec2 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58>
@@ -959,7 +952,6 @@ define void @load_i16_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61>
   %strided.vec6 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62>
   %strided.vec7 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63>
-
   store <8 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i16> %strided.vec2, ptr %out.vec2, align 64
@@ -968,7 +960,6 @@ define void @load_i16_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <8 x i16> %strided.vec5, ptr %out.vec5, align 64
   store <8 x i16> %strided.vec6, ptr %out.vec6, align 64
   store <8 x i16> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -1933,7 +1924,6 @@ define void @load_i16_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <128 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120>
   %strided.vec1 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121>
   %strided.vec2 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122>
@@ -1942,7 +1932,6 @@ define void @load_i16_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   %strided.vec5 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125>
   %strided.vec6 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126>
   %strided.vec7 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127>
-
   store <16 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i16> %strided.vec2, ptr %out.vec2, align 64
@@ -1951,7 +1940,6 @@ define void @load_i16_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   store <16 x i16> %strided.vec5, ptr %out.vec5, align 64
   store <16 x i16> %strided.vec6, ptr %out.vec6, align 64
   store <16 x i16> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -4206,7 +4194,6 @@ define void @load_i16_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <256 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120, i32 128, i32 136, i32 144, i32 152, i32 160, i32 168, i32 176, i32 184, i32 192, i32 200, i32 208, i32 216, i32 224, i32 232, i32 240, i32 248>
   %strided.vec1 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121, i32 129, i32 137, i32 145, i32 153, i32 161, i32 169, i32 177, i32 185, i32 193, i32 201, i32 209, i32 217, i32 225, i32 233, i32 241, i32 249>
   %strided.vec2 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122, i32 130, i32 138, i32 146, i32 154, i32 162, i32 170, i32 178, i32 186, i32 194, i32 202, i32 210, i32 218, i32 226, i32 234, i32 242, i32 250>
@@ -4215,7 +4202,6 @@ define void @load_i16_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   %strided.vec5 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125, i32 133, i32 141, i32 149, i32 157, i32 165, i32 173, i32 181, i32 189, i32 197, i32 205, i32 213, i32 221, i32 229, i32 237, i32 245, i32 253>
   %strided.vec6 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126, i32 134, i32 142, i32 150, i32 158, i32 166, i32 174, i32 182, i32 190, i32 198, i32 206, i32 214, i32 222, i32 230, i32 238, i32 246, i32 254>
   %strided.vec7 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127, i32 135, i32 143, i32 151, i32 159, i32 167, i32 175, i32 183, i32 191, i32 199, i32 207, i32 215, i32 223, i32 231, i32 239, i32 247, i32 255>
-
   store <32 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i16> %strided.vec2, ptr %out.vec2, align 64
@@ -4224,7 +4210,6 @@ define void @load_i16_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   store <32 x i16> %strided.vec5, ptr %out.vec5, align 64
   store <32 x i16> %strided.vec6, ptr %out.vec6, align 64
   store <32 x i16> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -9273,7 +9258,6 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <512 x i16>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120, i32 128, i32 136, i32 144, i32 152, i32 160, i32 168, i32 176, i32 184, i32 192, i32 200, i32 208, i32 216, i32 224, i32 232, i32 240, i32 248, i32 256, i32 264, i32 272, i32 280, i32 288, i32 296, i32 304, i32 312, i32 320, i32 328, i32 336, i32 344, i32 352, i32 360, i32 368, i32 376, i32 384, i32 392, i32 400, i32 408, i32 416, i32 424, i32 432, i32 440, i32 448, i32 456, i32 464, i32 472, i32 480, i32 488, i32 496, i32 504>
   %strided.vec1 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121, i32 129, i32 137, i32 145, i32 153, i32 161, i32 169, i32 177, i32 185, i32 193, i32 201, i32 209, i32 217, i32 225, i32 233, i32 241, i32 249, i32 257, i32 265, i32 273, i32 281, i32 289, i32 297, i32 305, i32 313, i32 321, i32 329, i32 337, i32 345, i32 353, i32 361, i32 369, i32 377, i32 385, i32 393, i32 401, i32 409, i32 417, i32 425, i32 433, i32 441, i32 449, i32 457, i32 465, i32 473, i32 481, i32 489, i32 497, i32 505>
   %strided.vec2 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122, i32 130, i32 138, i32 146, i32 154, i32 162, i32 170, i32 178, i32 186, i32 194, i32 202, i32 210, i32 218, i32 226, i32 234, i32 242, i32 250, i32 258, i32 266, i32 274, i32 282, i32 290, i32 298, i32 306, i32 314, i32 322, i32 330, i32 338, i32 346, i32 354, i32 362, i32 370, i32 378, i32 386, i32 394, i32 402, i32 410, i32 418, i32 426, i32 434, i32 442, i32 450, i32 458, i32 466, i32 474, i32 482, i32 490, i32 498, i32 506>
@@ -9282,7 +9266,6 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   %strided.vec5 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125, i32 133, i32 141, i32 149, i32 157, i32 165, i32 173, i32 181, i32 189, i32 197, i32 205, i32 213, i32 221, i32 229, i32 237, i32 245, i32 253, i32 261, i32 269, i32 277, i32 285, i32 293, i32 301, i32 309, i32 317, i32 325, i32 333, i32 341, i32 349, i32 357, i32 365, i32 373, i32 381, i32 389, i32 397, i32 405, i32 413, i32 421, i32 429, i32 437, i32 445, i32 453, i32 461, i32 469, i32 477, i32 485, i32 493, i32 501, i32 509>
   %strided.vec6 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126, i32 134, i32 142, i32 150, i32 158, i32 166, i32 174, i32 182, i32 190, i32 198, i32 206, i32 214, i32 222, i32 230, i32 238, i32 246, i32 254, i32 262, i32 270, i32 278, i32 286, i32 294, i32 302, i32 310, i32 318, i32 326, i32 334, i32 342, i32 350, i32 358, i32 366, i32 374, i32 382, i32 390, i32 398, i32 406, i32 414, i32 422, i32 430, i32 438, i32 446, i32 454, i32 462, i32 470, i32 478, i32 486, i32 494, i32 502, i32 510>
   %strided.vec7 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127, i32 135, i32 143, i32 151, i32 159, i32 167, i32 175, i32 183, i32 191, i32 199, i32 207, i32 215, i32 223, i32 231, i32 239, i32 247, i32 255, i32 263, i32 271, i32 279, i32 287, i32 295, i32 303, i32 311, i32 319, i32 327, i32 335, i32 343, i32 351, i32 359, i32 367, i32 375, i32 383, i32 391, i32 399, i32 407, i32 415, i32 423, i32 431, i32 439, i32 447, i32 455, i32 463, i32 471, i32 479, i32 487, i32 495, i32 503, i32 511>
-
   store <64 x i16> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i16> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i16> %strided.vec2, ptr %out.vec2, align 64
@@ -9291,7 +9274,6 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   store <64 x i16> %strided.vec5, ptr %out.vec5, align 64
   store <64 x i16> %strided.vec6, ptr %out.vec6, align 64
   store <64 x i16> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-2.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-2.ll
index 583d586..40514db 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-2.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-2.ll
@@ -34,13 +34,10 @@ define void @load_i32_stride2_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nou
 ; AVX-NEXT:    vmovlps %xmm0, (%rdx)
 ; AVX-NEXT:    retq
   %wide.vec = load <4 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <4 x i32> %wide.vec, <4 x i32> poison, <2 x i32> <i32 0, i32 2>
   %strided.vec1 = shufflevector <4 x i32> %wide.vec, <4 x i32> poison, <2 x i32> <i32 1, i32 3>
-
   store <2 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i32> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -76,13 +73,10 @@ define void @load_i32_stride2_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nou
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <8 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <8 x i32> %wide.vec, <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
   %strided.vec1 = shufflevector <8 x i32> %wide.vec, <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
-
   store <4 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i32> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -174,13 +168,10 @@ define void @load_i32_stride2_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nou
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <16 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <16 x i32> %wide.vec, <16 x i32> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %strided.vec1 = shufflevector <16 x i32> %wide.vec, <16 x i32> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
-
   store <8 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i32> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -270,13 +261,10 @@ define void @load_i32_stride2_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) no
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <32 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <32 x i32> %wide.vec, <32 x i32> poison, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %strided.vec1 = shufflevector <32 x i32> %wide.vec, <32 x i32> poison, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
-
   store <16 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i32> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -433,13 +421,10 @@ define void @load_i32_stride2_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) no
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <64 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <64 x i32> %wide.vec, <64 x i32> poison, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46, i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62>
   %strided.vec1 = shufflevector <64 x i32> %wide.vec, <64 x i32> poison, <32 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31, i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47, i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63>
-
   store <32 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i32> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -764,13 +749,10 @@ define void @load_i32_stride2_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) no
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <128 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <64 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46, i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62, i32 64, i32 66, i32 68, i32 70, i32 72, i32 74, i32 76, i32 78, i32 80, i32 82, i32 84, i32 86, i32 88, i32 90, i32 92, i32 94, i32 96, i32 98, i32 100, i32 102, i32 104, i32 106, i32 108, i32 110, i32 112, i32 114, i32 116, i32 118, i32 120, i32 122, i32 124, i32 126>
   %strided.vec1 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <64 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31, i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47, i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63, i32 65, i32 67, i32 69, i32 71, i32 73, i32 75, i32 77, i32 79, i32 81, i32 83, i32 85, i32 87, i32 89, i32 91, i32 93, i32 95, i32 97, i32 99, i32 101, i32 103, i32 105, i32 107, i32 109, i32 111, i32 113, i32 115, i32 117, i32 119, i32 121, i32 123, i32 125, i32 127>
-
   store <64 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i32> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
index c6d50fe..9779b49 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
@@ -115,15 +115,12 @@ define void @load_i32_stride3_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vmovq %xmm0, (%rcx)
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <6 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <6 x i32> %wide.vec, <6 x i32> poison, <2 x i32> <i32 0, i32 3>
   %strided.vec1 = shufflevector <6 x i32> %wide.vec, <6 x i32> poison, <2 x i32> <i32 1, i32 4>
   %strided.vec2 = shufflevector <6 x i32> %wide.vec, <6 x i32> poison, <2 x i32> <i32 2, i32 5>
-
   store <2 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i32> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -204,15 +201,12 @@ define void @load_i32_stride3_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <12 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
   %strided.vec1 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
   %strided.vec2 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
-
   store <4 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i32> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -392,15 +386,12 @@ define void @load_i32_stride3_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <24 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
   %strided.vec1 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
   %strided.vec2 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
-
   store <8 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i32> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -727,15 +718,12 @@ define void @load_i32_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <48 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
   %strided.vec1 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
   %strided.vec2 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
-
   store <16 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i32> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -1449,15 +1437,12 @@ define void @load_i32_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <96 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <32 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45, i32 48, i32 51, i32 54, i32 57, i32 60, i32 63, i32 66, i32 69, i32 72, i32 75, i32 78, i32 81, i32 84, i32 87, i32 90, i32 93>
   %strided.vec1 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <32 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46, i32 49, i32 52, i32 55, i32 58, i32 61, i32 64, i32 67, i32 70, i32 73, i32 76, i32 79, i32 82, i32 85, i32 88, i32 91, i32 94>
   %strided.vec2 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <32 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47, i32 50, i32 53, i32 56, i32 59, i32 62, i32 65, i32 68, i32 71, i32 74, i32 77, i32 80, i32 83, i32 86, i32 89, i32 92, i32 95>
-
   store <32 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i32> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -3037,15 +3022,12 @@ define void @load_i32_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <192 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <64 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45, i32 48, i32 51, i32 54, i32 57, i32 60, i32 63, i32 66, i32 69, i32 72, i32 75, i32 78, i32 81, i32 84, i32 87, i32 90, i32 93, i32 96, i32 99, i32 102, i32 105, i32 108, i32 111, i32 114, i32 117, i32 120, i32 123, i32 126, i32 129, i32 132, i32 135, i32 138, i32 141, i32 144, i32 147, i32 150, i32 153, i32 156, i32 159, i32 162, i32 165, i32 168, i32 171, i32 174, i32 177, i32 180, i32 183, i32 186, i32 189>
   %strided.vec1 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <64 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46, i32 49, i32 52, i32 55, i32 58, i32 61, i32 64, i32 67, i32 70, i32 73, i32 76, i32 79, i32 82, i32 85, i32 88, i32 91, i32 94, i32 97, i32 100, i32 103, i32 106, i32 109, i32 112, i32 115, i32 118, i32 121, i32 124, i32 127, i32 130, i32 133, i32 136, i32 139, i32 142, i32 145, i32 148, i32 151, i32 154, i32 157, i32 160, i32 163, i32 166, i32 169, i32 172, i32 175, i32 178, i32 181, i32 184, i32 187, i32 190>
   %strided.vec2 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <64 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47, i32 50, i32 53, i32 56, i32 59, i32 62, i32 65, i32 68, i32 71, i32 74, i32 77, i32 80, i32 83, i32 86, i32 89, i32 92, i32 95, i32 98, i32 101, i32 104, i32 107, i32 110, i32 113, i32 116, i32 119, i32 122, i32 125, i32 128, i32 131, i32 134, i32 137, i32 140, i32 143, i32 146, i32 149, i32 152, i32 155, i32 158, i32 161, i32 164, i32 167, i32 170, i32 173, i32 176, i32 179, i32 182, i32 185, i32 188, i32 191>
-
   store <64 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i32> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
index ed0e561..d7ba515 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
@@ -115,17 +115,14 @@ define void @load_i32_stride4_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vpextrq $1, %xmm0, (%r8)
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <8 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <8 x i32> %wide.vec, <8 x i32> poison, <2 x i32> <i32 0, i32 4>
   %strided.vec1 = shufflevector <8 x i32> %wide.vec, <8 x i32> poison, <2 x i32> <i32 1, i32 5>
   %strided.vec2 = shufflevector <8 x i32> %wide.vec, <8 x i32> poison, <2 x i32> <i32 2, i32 6>
   %strided.vec3 = shufflevector <8 x i32> %wide.vec, <8 x i32> poison, <2 x i32> <i32 3, i32 7>
-
   store <2 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <2 x i32> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -228,17 +225,14 @@ define void @load_i32_stride4_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <16 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <16 x i32> %wide.vec, <16 x i32> poison, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
   %strided.vec1 = shufflevector <16 x i32> %wide.vec, <16 x i32> poison, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
   %strided.vec2 = shufflevector <16 x i32> %wide.vec, <16 x i32> poison, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
   %strided.vec3 = shufflevector <16 x i32> %wide.vec, <16 x i32> poison, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
-
   store <4 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <4 x i32> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -404,17 +398,14 @@ define void @load_i32_stride4_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <32 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <32 x i32> %wide.vec, <32 x i32> poison, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
   %strided.vec1 = shufflevector <32 x i32> %wide.vec, <32 x i32> poison, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
   %strided.vec2 = shufflevector <32 x i32> %wide.vec, <32 x i32> poison, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
   %strided.vec3 = shufflevector <32 x i32> %wide.vec, <32 x i32> poison, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
-
   store <8 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <8 x i32> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -804,17 +795,14 @@ define void @load_i32_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <64 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <64 x i32> %wide.vec, <64 x i32> poison, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
   %strided.vec1 = shufflevector <64 x i32> %wide.vec, <64 x i32> poison, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
   %strided.vec2 = shufflevector <64 x i32> %wide.vec, <64 x i32> poison, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
   %strided.vec3 = shufflevector <64 x i32> %wide.vec, <64 x i32> poison, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
-
   store <16 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <16 x i32> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -1666,17 +1654,1809 @@ define void @load_i32_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <128 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <32 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124>
   %strided.vec1 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <32 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125>
   %strided.vec2 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <32 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126>
   %strided.vec3 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <32 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127>
-
   store <32 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <32 x i32> %strided.vec3, ptr %out.vec3, align 64
+  ret void
+}
 
+define void @load_i32_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
+; SSE-LABEL: load_i32_stride4_vf64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    subq $1224, %rsp # imm = 0x4C8
+; SSE-NEXT:    movaps 144(%rdi), %xmm4
+; SSE-NEXT:    movaps 176(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 160(%rdi), %xmm6
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 208(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 192(%rdi), %xmm8
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 240(%rdi), %xmm9
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 224(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 64(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 112(%rdi), %xmm12
+; SSE-NEXT:    movaps 96(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm8, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm0
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT:    movaps 128(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm4, %xmm8
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 368(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 352(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 336(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 320(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 304(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 288(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movaps 272(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 256(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    movaps %xmm2, %xmm11
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 496(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 480(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 464(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 448(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 432(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 416(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 400(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 384(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 624(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 608(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 592(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 576(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 560(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 544(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 528(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 512(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 752(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 736(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 720(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 704(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 688(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 672(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 656(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 640(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 880(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 864(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 848(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 832(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 816(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 800(%rdi), %xmm13
+; SSE-NEXT:    movaps %xmm13, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 784(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 768(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1008(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 992(%rdi), %xmm15
+; SSE-NEXT:    movaps %xmm15, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 976(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 960(%rdi), %xmm14
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm14, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm0[1]
+; SSE-NEXT:    movaps 944(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 928(%rdi), %xmm9
+; SSE-NEXT:    movaps %xmm9, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movaps 912(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 896(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm5, %xmm0
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm2[1]
+; SSE-NEXT:    movaps 32(%rdi), %xmm6
+; SSE-NEXT:    movaps 48(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm4
+; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    movaps (%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm10, %xmm0
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm4[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm8[2],xmm3[3],xmm8[3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = xmm11[2],mem[2],xmm11[3],mem[3]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = xmm11[2],mem[2],xmm11[3],mem[3]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = xmm11[2],mem[2],xmm11[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = xmm13[2],mem[2],xmm13[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[2],mem[2],xmm9[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm12, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, %xmm12
+; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm4[0]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm8, %xmm1
+; SSE-NEXT:    movaps %xmm8, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm7[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm7[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm2
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm11, %xmm0
+; SSE-NEXT:    movaps %xmm11, %xmm12
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm1[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm11
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm11 = xmm11[0],xmm1[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm1[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm1[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm1
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, %xmm0
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm13[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm13[1]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, %xmm13
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm13 = xmm13[0],xmm2[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, %xmm8
+; SSE-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm9[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm2, %xmm9
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm15[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm15[1]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm6[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm6[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 224(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 160(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 240(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 176(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 192(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 128(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 208(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 144(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 16(%rsi)
+; SSE-NEXT:    movaps %xmm5, 224(%rdx)
+; SSE-NEXT:    movaps %xmm14, 240(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 192(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 208(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 160(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 176(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 128(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 144(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 48(%rdx)
+; SSE-NEXT:    movaps %xmm10, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 16(%rdx)
+; SSE-NEXT:    movaps %xmm3, 240(%rcx)
+; SSE-NEXT:    movaps %xmm8, 224(%rcx)
+; SSE-NEXT:    movaps %xmm13, 208(%rcx)
+; SSE-NEXT:    movaps %xmm0, 192(%rcx)
+; SSE-NEXT:    movaps %xmm1, 176(%rcx)
+; SSE-NEXT:    movaps %xmm4, 160(%rcx)
+; SSE-NEXT:    movaps %xmm7, 144(%rcx)
+; SSE-NEXT:    movaps %xmm11, 128(%rcx)
+; SSE-NEXT:    movaps %xmm12, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps %xmm2, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r8)
+; SSE-NEXT:    movaps %xmm9, 224(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r8)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r8)
+; SSE-NEXT:    movaps %xmm15, (%r8)
+; SSE-NEXT:    addq $1224, %rsp # imm = 0x4C8
+; SSE-NEXT:    retq
+;
+; AVX1-ONLY-LABEL: load_i32_stride4_vf64:
+; AVX1-ONLY:       # %bb.0:
+; AVX1-ONLY-NEXT:    subq $2200, %rsp # imm = 0x898
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm14[0],ymm1[2],ymm14[2]
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, %ymm8
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm9[2,3,0,1]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm5[0],ymm9[0],ymm5[1],ymm9[1],ymm5[4],ymm9[4],ymm5[5],ymm9[5]
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, %ymm7
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 176(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm11[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm10
+; AVX1-ONLY-NEXT:    vmovaps 144(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, %xmm15
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm4[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm4[0],ymm0[0],ymm4[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm3[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 432(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps 400(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, %xmm6
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 688(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps 656(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 944(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps 912(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 304(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps 272(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 560(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps 528(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 816(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps 784(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm12[0,1],xmm1[2,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm13 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[0,1],ymm12[2,0],ymm13[4,5],ymm12[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm13[0,1],xmm0[2,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm14[0],ymm8[0],ymm14[1],ymm8[1],ymm14[4],ymm8[4],ymm14[5],ymm8[5]
+; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[1,0],ymm7[1,0],ymm9[5,4],ymm7[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, %xmm7
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = xmm15[1],xmm5[1],zero,zero
+; AVX1-ONLY-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps %xmm10, %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm10[0],xmm11[0],xmm10[1],xmm11[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm4[0],ymm8[0],ymm4[1],ymm8[1],ymm4[4],ymm8[4],ymm4[5],ymm8[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[1,0],ymm5[1,0],ymm10[5,4],ymm5[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = xmm6[1],xmm15[1],zero,zero
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm6[0],xmm12[0],xmm6[1],xmm12[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0],xmm1[1],zero,zero
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm13[0],mem[0],xmm13[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0],xmm1[1],zero,zero
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm13[0],mem[0],xmm13[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0],xmm1[1],zero,zero
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm13[0],mem[0],xmm13[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0],xmm1[1],zero,zero
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm13[0],mem[0],xmm13[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0],xmm1[1],zero,zero
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm13[0],mem[0],xmm13[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = mem[0],xmm13[1],zero,zero
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm14[1],ymm2[3],ymm14[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm14[2],ymm9[2],ymm14[3],ymm9[3],ymm14[6],ymm9[6],ymm14[7],ymm9[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm7[2],mem[2],xmm7[3],mem[3]
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm13 = zero,zero,xmm3[2],xmm11[2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm8[1],ymm4[1],ymm8[3],ymm4[3]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm10[2],ymm5[3],ymm10[3],ymm5[6],ymm10[6],ymm5[7],ymm10[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm1[2],xmm15[2],xmm1[3],xmm15[3]
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm13 = zero,zero,xmm6[2],xmm12[2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm4[1],ymm11[1],ymm4[3],ymm11[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm9[2],ymm5[3],ymm9[3],ymm5[6],ymm9[6],ymm5[7],ymm9[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm15[2],xmm12[2],xmm15[3],xmm12[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm13 = zero,zero,xmm2[2],xmm3[2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm6[1],ymm8[1],ymm6[3],ymm8[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm7[2],ymm10[2],ymm7[3],ymm10[3],ymm7[6],ymm10[6],ymm7[7],ymm10[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = zero,zero,xmm13[2],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = zero,zero,xmm13[2],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps (%rsp), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = zero,zero,xmm13[2],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = zero,zero,xmm13[2],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = zero,zero,xmm1[2],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm13[2],mem[2],xmm13[3],mem[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm14[3,0],ymm1[7,4],ymm14[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm13[3,0],mem[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm13[2,0],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm13[3,0],mem[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm13[2,0],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm11[2],ymm4[2],ymm11[3],ymm4[3],ymm11[6],ymm4[6],ymm11[7],ymm4[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[3,0],ymm5[3,0],ymm9[7,4],ymm5[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm12[3,0],xmm15[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm13[2,0],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm6[2],ymm8[3],ymm6[3],ymm8[6],ymm6[6],ymm8[7],ymm6[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[3,0],ymm7[3,0],ymm10[7,4],ymm7[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[2,0],ymm1[2,3],ymm10[6,4],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm2[3,0],mem[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm11[2,0],xmm10[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm2[2],mem[2],ymm2[3],mem[3],ymm2[6],mem[6],ymm2[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm9 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm9 = ymm2[3,0],mem[3,0],ymm2[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm9[2,0],ymm6[2,3],ymm9[6,4],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm9 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm2[3,0],mem[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm10[2,0],xmm9[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = ymm2[2],mem[2],ymm2[3],mem[3],ymm2[6],mem[6],ymm2[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm2[3,0],mem[3,0],ymm2[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm8[2,0],ymm4[2,3],ymm8[6,4],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm9 = xmm2[3,0],mem[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm9[2,0],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm8[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm2[2],mem[2],ymm2[3],mem[3],ymm2[6],mem[6],ymm2[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = ymm3[3,0],mem[3,0],ymm3[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[2,0],ymm2[2,3],ymm7[6,4],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = xmm3[3,0],mem[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm8[2,0],xmm7[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[2],mem[2],ymm3[3],mem[3],ymm3[6],mem[6],ymm3[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm5 = ymm5[3,0],mem[3,0],ymm5[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm5[2,0],ymm3[2,3],ymm5[6,4],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = xmm7[3,0],mem[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm7[2,0],xmm5[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 192(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 128(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, (%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 224(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 160(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 192(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 128(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, (%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 224(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 160(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, (%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 192(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 128(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm6, 64(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm14, 32(%r8)
+; AVX1-ONLY-NEXT:    addq $2200, %rsp # imm = 0x898
+; AVX1-ONLY-NEXT:    vzeroupper
+; AVX1-ONLY-NEXT:    retq
+;
+; AVX2-ONLY-LABEL: load_i32_stride4_vf64:
+; AVX2-ONLY:       # %bb.0:
+; AVX2-ONLY-NEXT:    subq $1960, %rsp # imm = 0x7A8
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vbroadcastsd {{.*#+}} ymm6 = [17179869184,17179869184,17179869184,17179869184]
+; AVX2-ONLY-NEXT:    vpermps %ymm1, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, %ymm5
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm2, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, %ymm10
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 144(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vmovaps {{.*#+}} xmm7 = <u,u,0,4>
+; AVX2-ONLY-NEXT:    vpermps %ymm9, %ymm7, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm4, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm3, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 400(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vpermps %ymm8, %ymm7, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 704(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 736(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpermps %ymm1, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, %ymm9
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm2, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 672(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 656(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 640(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vpermps %ymm3, %ymm7, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 960(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovaps 992(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm0, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vpermps %ymm2, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, %ymm14
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 928(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 912(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 896(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vpermps %ymm3, %ymm7, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpermps %ymm1, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, %ymm3
+; AVX2-ONLY-NEXT:    vpermps %ymm2, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 272(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vpermps %ymm8, %ymm7, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 576(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 608(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpermps %ymm1, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, %ymm11
+; AVX2-ONLY-NEXT:    vpermps %ymm2, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 544(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 528(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 512(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vpermps %ymm8, %ymm7, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 832(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 864(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpermps %ymm1, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, %ymm13
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm2, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 800(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 784(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 768(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vpermps %ymm8, %ymm7, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpermps %ymm1, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, %ymm15
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm2, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, %ymm12
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm1, %ymm7, %ymm1
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 16(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd {{.*#+}} ymm6 = [21474836481,21474836481,21474836481,21474836481]
+; AVX2-ONLY-NEXT:    vpermps %ymm5, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps %ymm10, %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm10, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 176(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
+; AVX2-ONLY-NEXT:    vmovaps {{.*#+}} xmm7 = <1,5,u,u>
+; AVX2-ONLY-NEXT:    vpermps %ymm5, %ymm7, %ymm10
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm4, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm8, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 432(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
+; AVX2-ONLY-NEXT:    vpermps %ymm4, %ymm7, %ymm10
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm9, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm5, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 640(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 688(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 672(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
+; AVX2-ONLY-NEXT:    vpermps %ymm4, %ymm7, %ymm10
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm4, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vpermps %ymm14, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 896(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 944(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vmovaps %xmm10, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 928(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
+; AVX2-ONLY-NEXT:    vpermps %ymm9, %ymm7, %ymm10
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm15, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vpermps %ymm12, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 48(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vpermps %ymm9, %ymm7, %ymm10
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm3, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps %ymm3, %ymm15
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm3, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 304(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
+; AVX2-ONLY-NEXT:    vpermps %ymm9, %ymm7, %ymm10
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm11, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps %ymm11, %ymm12
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm11, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 512(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 560(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 544(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
+; AVX2-ONLY-NEXT:    vpermps %ymm9, %ymm7, %ymm10
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm13, %ymm6, %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm13, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 768(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm1, %ymm7, %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 816(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 800(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd {{.*#+}} ymm0 = [25769803778,25769803778,25769803778,25769803778]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm9, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermps %ymm2, %ymm0, %ymm10
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{.*#+}} xmm2 = <u,u,2,6>
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm14, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermps %ymm8, %ymm0, %ymm7
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm8, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermps %ymm5, %ymm0, %ymm7
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm4, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm5, %ymm0, %ymm7
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm15, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermps %ymm3, %ymm0, %ymm7
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm12, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermps %ymm11, %ymm0, %ymm7
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm11, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermps %ymm13, %ymm0, %ymm7
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm6, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm4, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd {{.*#+}} ymm0 = [30064771075,30064771075,30064771075,30064771075]
+; AVX2-ONLY-NEXT:    vpermps %ymm9, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{.*#+}} xmm1 = <3,7,u,u>
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermps %ymm14, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermps %ymm8, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm9[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpermps %ymm5, %ymm0, %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps (%rsp), %xmm3, %xmm5 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm5 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermps %ymm6, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vpermps %ymm4, %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermps %ymm15, %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermps %ymm12, %ymm0, %ymm4
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm6[2],mem[2],xmm6[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm7[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermps %ymm11, %ymm0, %ymm4
+; AVX2-ONLY-NEXT:    vpermps %ymm13, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rcx)
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 192(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm8, 128(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm3, 64(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, (%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 224(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm9, 160(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm10, 96(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 32(%r8)
+; AVX2-ONLY-NEXT:    addq $1960, %rsp # imm = 0x7A8
+; AVX2-ONLY-NEXT:    vzeroupper
+; AVX2-ONLY-NEXT:    retq
+;
+; AVX512-LABEL: load_i32_stride4_vf64:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm0
+; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm1
+; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm2
+; AVX512-NEXT:    vmovdqa64 192(%rdi), %zmm3
+; AVX512-NEXT:    vmovdqa64 832(%rdi), %zmm4
+; AVX512-NEXT:    vmovdqa64 768(%rdi), %zmm5
+; AVX512-NEXT:    vmovdqa64 960(%rdi), %zmm10
+; AVX512-NEXT:    vmovdqa64 896(%rdi), %zmm7
+; AVX512-NEXT:    vmovdqa64 320(%rdi), %zmm14
+; AVX512-NEXT:    vmovdqa64 256(%rdi), %zmm15
+; AVX512-NEXT:    vmovdqa64 448(%rdi), %zmm17
+; AVX512-NEXT:    vmovdqa64 384(%rdi), %zmm16
+; AVX512-NEXT:    vmovdqa64 576(%rdi), %zmm8
+; AVX512-NEXT:    vmovdqa64 512(%rdi), %zmm9
+; AVX512-NEXT:    vmovdqa64 704(%rdi), %zmm12
+; AVX512-NEXT:    vmovdqa64 640(%rdi), %zmm11
+; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,4,8,12,16,20,24,28,0,4,8,12,16,20,24,28]
+; AVX512-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm6
+; AVX512-NEXT:    vpermt2d %zmm12, %zmm19, %zmm6
+; AVX512-NEXT:    vmovdqa64 %zmm9, %zmm13
+; AVX512-NEXT:    vpermt2d %zmm8, %zmm19, %zmm13
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm13[0,1,2,3],zmm6[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %zmm16, %zmm13
+; AVX512-NEXT:    vpermt2d %zmm17, %zmm19, %zmm13
+; AVX512-NEXT:    vmovdqa64 %zmm15, %zmm18
+; AVX512-NEXT:    vpermt2d %zmm14, %zmm19, %zmm18
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm18[0,1,2,3],zmm13[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm18
+; AVX512-NEXT:    vpermt2d %zmm10, %zmm19, %zmm18
+; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm20
+; AVX512-NEXT:    vpermt2d %zmm4, %zmm19, %zmm20
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm18 = zmm20[0,1,2,3],zmm18[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm20
+; AVX512-NEXT:    vpermt2d %zmm3, %zmm19, %zmm20
+; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm19
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm20[4,5,6,7]
+; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [1,5,9,13,17,21,25,29,1,5,9,13,17,21,25,29]
+; AVX512-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm20
+; AVX512-NEXT:    vpermt2d %zmm12, %zmm21, %zmm20
+; AVX512-NEXT:    vmovdqa64 %zmm9, %zmm22
+; AVX512-NEXT:    vpermt2d %zmm8, %zmm21, %zmm22
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm20 = zmm22[0,1,2,3],zmm20[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %zmm16, %zmm22
+; AVX512-NEXT:    vpermt2d %zmm17, %zmm21, %zmm22
+; AVX512-NEXT:    vmovdqa64 %zmm15, %zmm23
+; AVX512-NEXT:    vpermt2d %zmm14, %zmm21, %zmm23
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm22 = zmm23[0,1,2,3],zmm22[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm23
+; AVX512-NEXT:    vpermt2d %zmm10, %zmm21, %zmm23
+; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm24
+; AVX512-NEXT:    vpermt2d %zmm4, %zmm21, %zmm24
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm24[0,1,2,3],zmm23[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm24
+; AVX512-NEXT:    vpermt2d %zmm3, %zmm21, %zmm24
+; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm21
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm24[4,5,6,7]
+; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm24 = [2,6,10,14,18,22,26,30,2,6,10,14,18,22,26,30]
+; AVX512-NEXT:    # zmm24 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm25
+; AVX512-NEXT:    vpermt2d %zmm12, %zmm24, %zmm25
+; AVX512-NEXT:    vmovdqa64 %zmm9, %zmm26
+; AVX512-NEXT:    vpermt2d %zmm8, %zmm24, %zmm26
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm25 = zmm26[0,1,2,3],zmm25[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %zmm16, %zmm26
+; AVX512-NEXT:    vpermt2d %zmm17, %zmm24, %zmm26
+; AVX512-NEXT:    vmovdqa64 %zmm15, %zmm27
+; AVX512-NEXT:    vpermt2d %zmm14, %zmm24, %zmm27
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm27[0,1,2,3],zmm26[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm27
+; AVX512-NEXT:    vpermt2d %zmm10, %zmm24, %zmm27
+; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm28
+; AVX512-NEXT:    vpermt2d %zmm4, %zmm24, %zmm28
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm27 = zmm28[0,1,2,3],zmm27[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm28
+; AVX512-NEXT:    vpermt2d %zmm3, %zmm24, %zmm28
+; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm24
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm24 = zmm24[0,1,2,3],zmm28[4,5,6,7]
+; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm28 = [3,7,11,15,19,23,27,31,3,7,11,15,19,23,27,31]
+; AVX512-NEXT:    # zmm28 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vpermt2d %zmm17, %zmm28, %zmm16
+; AVX512-NEXT:    vpermt2d %zmm14, %zmm28, %zmm15
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm15[0,1,2,3],zmm16[4,5,6,7]
+; AVX512-NEXT:    vpermt2d %zmm10, %zmm28, %zmm7
+; AVX512-NEXT:    vpermt2d %zmm4, %zmm28, %zmm5
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
+; AVX512-NEXT:    vpermt2d %zmm12, %zmm28, %zmm11
+; AVX512-NEXT:    vpermt2d %zmm8, %zmm28, %zmm9
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm9[0,1,2,3],zmm11[4,5,6,7]
+; AVX512-NEXT:    vpermt2d %zmm3, %zmm28, %zmm2
+; AVX512-NEXT:    vpermt2d %zmm1, %zmm28, %zmm0
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %zmm18, 192(%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm19, (%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm6, 128(%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm23, 192(%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm21, (%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm22, 64(%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm20, 128(%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm27, 192(%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm24, (%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm26, 64(%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm5, 128(%r8)
+; AVX512-NEXT:    vmovdqa64 %zmm4, 192(%r8)
+; AVX512-NEXT:    vmovdqa64 %zmm0, (%r8)
+; AVX512-NEXT:    vmovdqa64 %zmm14, 64(%r8)
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
+  %wide.vec = load <256 x i32>, ptr %in.vec, align 64
+  %strided.vec0 = shufflevector <256 x i32> %wide.vec, <256 x i32> poison, <64 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124, i32 128, i32 132, i32 136, i32 140, i32 144, i32 148, i32 152, i32 156, i32 160, i32 164, i32 168, i32 172, i32 176, i32 180, i32 184, i32 188, i32 192, i32 196, i32 200, i32 204, i32 208, i32 212, i32 216, i32 220, i32 224, i32 228, i32 232, i32 236, i32 240, i32 244, i32 248, i32 252>
+  %strided.vec1 = shufflevector <256 x i32> %wide.vec, <256 x i32> poison, <64 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125, i32 129, i32 133, i32 137, i32 141, i32 145, i32 149, i32 153, i32 157, i32 161, i32 165, i32 169, i32 173, i32 177, i32 181, i32 185, i32 189, i32 193, i32 197, i32 201, i32 205, i32 209, i32 213, i32 217, i32 221, i32 225, i32 229, i32 233, i32 237, i32 241, i32 245, i32 249, i32 253>
+  %strided.vec2 = shufflevector <256 x i32> %wide.vec, <256 x i32> poison, <64 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126, i32 130, i32 134, i32 138, i32 142, i32 146, i32 150, i32 154, i32 158, i32 162, i32 166, i32 170, i32 174, i32 178, i32 182, i32 186, i32 190, i32 194, i32 198, i32 202, i32 206, i32 210, i32 214, i32 218, i32 222, i32 226, i32 230, i32 234, i32 238, i32 242, i32 246, i32 250, i32 254>
+  %strided.vec3 = shufflevector <256 x i32> %wide.vec, <256 x i32> poison, <64 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127, i32 131, i32 135, i32 139, i32 143, i32 147, i32 151, i32 155, i32 159, i32 163, i32 167, i32 171, i32 175, i32 179, i32 183, i32 187, i32 191, i32 195, i32 199, i32 203, i32 207, i32 211, i32 215, i32 219, i32 223, i32 227, i32 231, i32 235, i32 239, i32 243, i32 247, i32 251, i32 255>
+  store <64 x i32> %strided.vec0, ptr %out.vec0, align 64
+  store <64 x i32> %strided.vec1, ptr %out.vec1, align 64
+  store <64 x i32> %strided.vec2, ptr %out.vec2, align 64
+  store <64 x i32> %strided.vec3, ptr %out.vec3, align 64
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
index 606d1d8..536a9cc 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
@@ -160,19 +160,16 @@ define void @load_i32_stride5_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <10 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <10 x i32> %wide.vec, <10 x i32> poison, <2 x i32> <i32 0, i32 5>
   %strided.vec1 = shufflevector <10 x i32> %wide.vec, <10 x i32> poison, <2 x i32> <i32 1, i32 6>
   %strided.vec2 = shufflevector <10 x i32> %wide.vec, <10 x i32> poison, <2 x i32> <i32 2, i32 7>
   %strided.vec3 = shufflevector <10 x i32> %wide.vec, <10 x i32> poison, <2 x i32> <i32 3, i32 8>
   %strided.vec4 = shufflevector <10 x i32> %wide.vec, <10 x i32> poison, <2 x i32> <i32 4, i32 9>
-
   store <2 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <2 x i32> %strided.vec3, ptr %out.vec3, align 64
   store <2 x i32> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -439,19 +436,16 @@ define void @load_i32_stride5_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <20 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <20 x i32> %wide.vec, <20 x i32> poison, <4 x i32> <i32 0, i32 5, i32 10, i32 15>
   %strided.vec1 = shufflevector <20 x i32> %wide.vec, <20 x i32> poison, <4 x i32> <i32 1, i32 6, i32 11, i32 16>
   %strided.vec2 = shufflevector <20 x i32> %wide.vec, <20 x i32> poison, <4 x i32> <i32 2, i32 7, i32 12, i32 17>
   %strided.vec3 = shufflevector <20 x i32> %wide.vec, <20 x i32> poison, <4 x i32> <i32 3, i32 8, i32 13, i32 18>
   %strided.vec4 = shufflevector <20 x i32> %wide.vec, <20 x i32> poison, <4 x i32> <i32 4, i32 9, i32 14, i32 19>
-
   store <4 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <4 x i32> %strided.vec3, ptr %out.vec3, align 64
   store <4 x i32> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -701,19 +695,16 @@ define void @load_i32_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <40 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <40 x i32> %wide.vec, <40 x i32> poison, <8 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35>
   %strided.vec1 = shufflevector <40 x i32> %wide.vec, <40 x i32> poison, <8 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36>
   %strided.vec2 = shufflevector <40 x i32> %wide.vec, <40 x i32> poison, <8 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37>
   %strided.vec3 = shufflevector <40 x i32> %wide.vec, <40 x i32> poison, <8 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38>
   %strided.vec4 = shufflevector <40 x i32> %wide.vec, <40 x i32> poison, <8 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39>
-
   store <8 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <8 x i32> %strided.vec3, ptr %out.vec3, align 64
   store <8 x i32> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -1336,19 +1327,16 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <80 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <80 x i32> %wide.vec, <80 x i32> poison, <16 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75>
   %strided.vec1 = shufflevector <80 x i32> %wide.vec, <80 x i32> poison, <16 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76>
   %strided.vec2 = shufflevector <80 x i32> %wide.vec, <80 x i32> poison, <16 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77>
   %strided.vec3 = shufflevector <80 x i32> %wide.vec, <80 x i32> poison, <16 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78>
   %strided.vec4 = shufflevector <80 x i32> %wide.vec, <80 x i32> poison, <16 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79>
-
   store <16 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <16 x i32> %strided.vec3, ptr %out.vec3, align 64
   store <16 x i32> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -2671,19 +2659,2756 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <160 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <160 x i32> %wide.vec, <160 x i32> poison, <32 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75, i32 80, i32 85, i32 90, i32 95, i32 100, i32 105, i32 110, i32 115, i32 120, i32 125, i32 130, i32 135, i32 140, i32 145, i32 150, i32 155>
   %strided.vec1 = shufflevector <160 x i32> %wide.vec, <160 x i32> poison, <32 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76, i32 81, i32 86, i32 91, i32 96, i32 101, i32 106, i32 111, i32 116, i32 121, i32 126, i32 131, i32 136, i32 141, i32 146, i32 151, i32 156>
   %strided.vec2 = shufflevector <160 x i32> %wide.vec, <160 x i32> poison, <32 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77, i32 82, i32 87, i32 92, i32 97, i32 102, i32 107, i32 112, i32 117, i32 122, i32 127, i32 132, i32 137, i32 142, i32 147, i32 152, i32 157>
   %strided.vec3 = shufflevector <160 x i32> %wide.vec, <160 x i32> poison, <32 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78, i32 83, i32 88, i32 93, i32 98, i32 103, i32 108, i32 113, i32 118, i32 123, i32 128, i32 133, i32 138, i32 143, i32 148, i32 153, i32 158>
   %strided.vec4 = shufflevector <160 x i32> %wide.vec, <160 x i32> poison, <32 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79, i32 84, i32 89, i32 94, i32 99, i32 104, i32 109, i32 114, i32 119, i32 124, i32 129, i32 134, i32 139, i32 144, i32 149, i32 154, i32 159>
-
   store <32 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <32 x i32> %strided.vec3, ptr %out.vec3, align 64
   store <32 x i32> %strided.vec4, ptr %out.vec4, align 64
+  ret void
+}
 
+define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4) nounwind {
+; SSE-LABEL: load_i32_stride5_vf64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    subq $1928, %rsp # imm = 0x788
+; SSE-NEXT:    movdqa 768(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 752(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 720(%rdi), %xmm11
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 736(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 448(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 432(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 400(%rdi), %xmm12
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 416(%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 128(%rdi), %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 112(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 80(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 96(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm8[2],xmm0[3],xmm8[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1040(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1056(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 1088(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1072(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rdi), %xmm10
+; SSE-NEXT:    movdqa 16(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 32(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 48(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 320(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 336(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 368(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 352(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 640(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 656(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 688(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 672(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 960(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 976(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 1008(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 992(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 240(%rdi), %xmm6
+; SSE-NEXT:    movdqa 256(%rdi), %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 288(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 272(%rdi), %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 560(%rdi), %xmm15
+; SSE-NEXT:    movdqa 576(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 608(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 592(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 880(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 896(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 928(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 912(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1200(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1216(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 1248(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1232(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 160(%rdi), %xmm14
+; SSE-NEXT:    movdqa 176(%rdi), %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 208(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 192(%rdi), %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 480(%rdi), %xmm8
+; SSE-NEXT:    movdqa 496(%rdi), %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 528(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 512(%rdi), %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 800(%rdi), %xmm13
+; SSE-NEXT:    movdqa 816(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 848(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 832(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1120(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa 1136(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 1168(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1152(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 144(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 64(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 304(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 224(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 464(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 384(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 624(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 544(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 784(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 704(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 944(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 864(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 1104(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 1024(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 1264(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rsp), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movdqa 1184(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm1[2],xmm14[3],xmm1[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm14[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm1[2],xmm12[3],xmm1[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[0,0,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm12[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[0,0,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,0,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
+; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm12, %xmm4
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
+; SSE-NEXT:    movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm4
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
+; SSE-NEXT:    movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm11, %xmm4
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm13, %xmm4
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, (%rsp), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm9, %xmm13
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = xmm14[2],mem[2],xmm14[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm4[0],xmm14[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm4[0],xmm12[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = xmm11[2],mem[2],xmm11[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm4[0],xmm11[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm1[0],xmm10[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[2],mem[2],xmm9[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm3[0],xmm6[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm15[0],xmm2[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm15[0],xmm1[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm15[0],xmm0[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 224(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 160(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 240(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 176(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 192(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 128(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 208(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 144(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 224(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 240(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 192(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 208(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 160(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 176(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 128(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 144(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 240(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 224(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 208(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 192(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 176(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 160(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 144(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 128(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, (%rcx)
+; SSE-NEXT:    movapd %xmm13, 240(%r8)
+; SSE-NEXT:    movaps (%rsp), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 224(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 208(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 192(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 176(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 160(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 144(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 128(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 112(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, (%r8)
+; SSE-NEXT:    movapd %xmm0, 240(%r9)
+; SSE-NEXT:    movapd %xmm1, 224(%r9)
+; SSE-NEXT:    movapd %xmm2, 208(%r9)
+; SSE-NEXT:    movapd %xmm3, 192(%r9)
+; SSE-NEXT:    movapd %xmm4, 176(%r9)
+; SSE-NEXT:    movapd %xmm5, 160(%r9)
+; SSE-NEXT:    movapd %xmm6, 144(%r9)
+; SSE-NEXT:    movapd %xmm7, 128(%r9)
+; SSE-NEXT:    movapd %xmm8, 112(%r9)
+; SSE-NEXT:    movapd %xmm9, 96(%r9)
+; SSE-NEXT:    movapd %xmm10, 80(%r9)
+; SSE-NEXT:    movapd %xmm11, 64(%r9)
+; SSE-NEXT:    movapd %xmm12, 48(%r9)
+; SSE-NEXT:    movapd %xmm14, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r9)
+; SSE-NEXT:    addq $1928, %rsp # imm = 0x788
+; SSE-NEXT:    retq
+;
+; AVX1-ONLY-LABEL: load_i32_stride5_vf64:
+; AVX1-ONLY:       # %bb.0:
+; AVX1-ONLY-NEXT:    subq $2440, %rsp # imm = 0x988
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1],ymm3[2,3],ymm6[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 608(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 928(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %ymm13
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1248(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm15
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %ymm11
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm9[4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 768(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0],xmm2[1],xmm0[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vmovaps 1024(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1088(%rdi), %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1],ymm0[1,3],ymm1[6,5],ymm0[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1],ymm5[2,3],ymm0[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[3,0],ymm1[6,4],ymm2[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 304(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1],ymm4[1,3],ymm1[6,5],ymm4[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm5[2,3],ymm4[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[3,0],ymm1[6,4],ymm2[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 624(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1],ymm3[1,3],ymm1[6,5],ymm3[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm6[2,3],ymm3[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, %ymm6
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[3,0],ymm1[6,4],ymm2[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 944(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm12[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1],ymm13[1,3],ymm1[6,5],ymm13[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm13[0,1],ymm12[2,3],ymm13[4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT:    vmovaps %ymm13, %ymm3
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[3,0],ymm1[6,4],ymm2[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 1264(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm15[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1],ymm7[1,3],ymm1[6,5],ymm7[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm15[2,3],ymm7[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[3,0],ymm1[6,4],ymm2[7,4]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 144(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm14[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1],ymm11[1,3],ymm1[6,5],ymm11[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm11[0,1],ymm14[2,3],ymm11[4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[3,0],ymm1[6,4],ymm2[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 464(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm10[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovaps %ymm9, %ymm13
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1],ymm9[1,3],ymm1[6,5],ymm9[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm10[2,3],ymm9[4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[3,0],ymm1[6,4],ymm2[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 784(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1],ymm8[1,3],ymm1[6,5],ymm8[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0,1],ymm0[2,3],ymm8[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[3,0],ymm1[6,4],ymm2[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm0[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 1104(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm2[2,0],ymm0[7,4],ymm2[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,1],ymm1[6,4],ymm2[6,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[1,0],ymm2[0,0],ymm0[5,4],ymm2[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm4[2,0],ymm0[7,4],ymm4[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm4[2,1],ymm1[6,4],ymm4[6,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[1,0],ymm2[0,0],ymm0[5,4],ymm2[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm6[2,0],ymm0[7,4],ymm6[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm6[2,1],ymm1[6,4],ymm6[6,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[1,0],ymm14[0,0],ymm0[5,4],ymm14[4,4]
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm3[2,0],ymm0[7,4],ymm3[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm3[2,1],ymm1[6,4],ymm3[6,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[1,0],ymm2[0,0],ymm0[5,4],ymm2[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm8[2,0],ymm0[7,4],ymm8[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm8[2,1],ymm0[6,4],ymm8[6,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,0],ymm1[0,0],ymm2[5,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm7[2,0],ymm0[7,4],ymm7[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm7[2,1],ymm0[6,4],ymm7[6,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm4[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm4[1,0],ymm10[0,0],ymm4[5,4],ymm10[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm11[2,0],ymm0[7,4],ymm11[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm11[2,1],ymm1[6,4],ymm11[6,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[1,0],ymm12[0,0],ymm6[5,4],ymm12[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[3,0],ymm9[2,0],ymm0[7,4],ymm9[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm9[2,1],ymm2[6,4],ymm9[6,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm5 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],mem[2],xmm5[3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm5 = xmm5[1,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm2[3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm8[1,0],ymm3[0,0],ymm8[5,4],ymm3[4,4]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm9[0,0],ymm1[3,0],ymm9[4,4],ymm1[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm1[2,0],mem[1,0],ymm1[6,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm2[3,0],ymm1[4,4],ymm2[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm2[2,2],ymm15[6,4],ymm2[6,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm1[2,0],mem[1,0],ymm1[6,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm2[0,0],ymm1[3,0],ymm2[4,4],ymm1[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[2,0],ymm14[1,0],ymm1[6,4],ymm14[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,0],ymm1[3,0],ymm15[4,4],ymm1[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm1[2,0],mem[1,0],ymm1[6,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm14[0,0],ymm1[3,0],ymm14[4,4],ymm1[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm1[2,0],mem[1,0],ymm1[6,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm5[0,0],ymm13[3,0],ymm5[4,4],ymm13[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm13[2,2],ymm15[6,4],ymm13[6,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm8[2,0],ymm3[1,0],ymm8[6,4],ymm3[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm2[0,0],ymm11[3,0],ymm2[4,4],ymm11[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm11[2,2],ymm15[6,4],ymm11[6,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm6[2,0],ymm12[1,0],ymm6[6,4],ymm12[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm7[3,0],ymm1[4,4],ymm7[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm7[2,2],ymm15[6,4],ymm7[6,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[2,0],ymm10[1,0],ymm4[6,4],ymm10[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3,4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm9[1],ymm3[2,3,4],ymm9[5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0],mem[1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = mem[0,1,2,3,4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm2[1],ymm4[2,3,4],ymm2[5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5,6],ymm6[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],mem[1],xmm6[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = mem[0,1,2,3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm5[1],ymm6[2,3,4],ymm5[5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5,6],ymm8[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],mem[1],xmm6[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $128, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm6[0,1,2,3,4,5,6],mem[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm14[1],ymm6[2,3,4],ymm14[5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = mem[0,1,2,3,4,5,6],ymm7[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0],mem[1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3,4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $128, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm8[0,1,2,3,4,5,6],mem[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm15, (%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 224(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm6, 192(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 160(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 96(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%r9)
+; AVX1-ONLY-NEXT:    addq $2440, %rsp # imm = 0x988
+; AVX1-ONLY-NEXT:    vzeroupper
+; AVX1-ONLY-NEXT:    retq
+;
+; AVX2-ONLY-LABEL: load_i32_stride5_vf64:
+; AVX2-ONLY:       # %bb.0:
+; AVX2-ONLY-NEXT:    subq $2120, %rsp # imm = 0x848
+; AVX2-ONLY-NEXT:    vmovdqa 864(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 896(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 832(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 800(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 544(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 576(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovdqa 512(%rdi), %ymm10
+; AVX2-ONLY-NEXT:    vmovdqa 480(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 224(%rdi), %ymm12
+; AVX2-ONLY-NEXT:    vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 256(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 192(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 160(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa {{.*#+}} xmm0 = [0,5,2,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vinserti128 $1, 288(%rdi), %ymm1, %ymm2
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[0,1,0,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4],ymm3[5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6],ymm2[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm10, %ymm14
+; AVX2-ONLY-NEXT:    vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm9[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm9, %ymm13
+; AVX2-ONLY-NEXT:    vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4],ymm2[5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, 608(%rdi), %ymm1, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm6[2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm5[0,1,0,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4],ymm2[5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, 928(%rdi), %ymm1, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1152(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqa 1120(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm3, %ymm9
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa %ymm2, %ymm8
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vmovdqa 1184(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1216(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm4[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm4, %ymm11
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, 1248(%rdi), %ymm1, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 352(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 320(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vmovdqa 384(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 416(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, 448(%rdi), %ymm1, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 672(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 640(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vmovdqa 704(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 736(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, 768(%rdi), %ymm1, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 992(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 960(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vmovdqa 1024(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1056(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, 1088(%rdi), %ymm1, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa (%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovdqa 32(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm15[2,3],ymm6[4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vmovdqa 64(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqa 96(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm5[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4],ymm1[5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, 128(%rdi), %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa {{.*#+}} xmm0 = <1,6,3,u>
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm12[0,1],ymm7[2,3],ymm12[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm10[0,1],mem[2,3],ymm10[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [5,2,7,0,5,2,7,0]
+; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 304(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm14[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 624(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 944(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm9[0,1],ymm8[2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm11[2,3],ymm9[4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 1264(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1],ymm15[2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1],ymm5[2,3],ymm4[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 144(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm11[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu (%rsp), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm3[0,1],mem[2,3],ymm3[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 464(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 784(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1],ymm8[2,3],ymm13[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm2, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm2, %ymm1, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 1104(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa {{.*#+}} xmm2 = <2,7,4,u>
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm7[4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm0, %ymm2, %ymm0
+; AVX2-ONLY-NEXT:    vinserti128 $1, 256(%rdi), %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vmovdqa %ymm10, %ymm4
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm1[3],ymm10[4,5,6],ymm1[7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 288(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq {{.*#+}} ymm0 = [21474836480,21474836480,21474836480,21474836480]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm10[0,1,2,3],mem[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm2, %ymm1
+; AVX2-ONLY-NEXT:    vinserti128 $1, 576(%rdi), %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm14[0,1,2],ymm3[3],ymm14[4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 608(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm2, %ymm1
+; AVX2-ONLY-NEXT:    vinserti128 $1, 896(%rdi), %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3],ymm5[4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 928(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm2, %ymm1
+; AVX2-ONLY-NEXT:    vinserti128 $1, 1216(%rdi), %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm9[0,1,2],ymm3[3],ymm9[4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1248(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm13[0,1,2,3],ymm8[4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm2, %ymm1
+; AVX2-ONLY-NEXT:    vinserti128 $1, 1056(%rdi), %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm15[0,1,2],ymm3[3],ymm15[4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1088(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vpermd %ymm8, %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm2, %ymm1
+; AVX2-ONLY-NEXT:    vinserti128 $1, 96(%rdi), %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd $119, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,2],ymm3[3],mem[4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 128(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm0, %ymm15
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,2,3],ymm11[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm2, %ymm3
+; AVX2-ONLY-NEXT:    vinserti128 $1, 416(%rdi), %ymm0, %ymm15
+; AVX2-ONLY-NEXT:    vpblendd $119, (%rsp), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = mem[0,1,2],ymm15[3],mem[4,5,6],ymm15[7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm15 = ymm15[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm15 = ymm3[0,1,2],ymm15[3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 448(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm0, %ymm13
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm15[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm11[0,1,2,3],ymm9[4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm13, %ymm2, %ymm2
+; AVX2-ONLY-NEXT:    vinserti128 $1, 736(%rdi), %ymm0, %ymm13
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3],ymm14[4,5,6],ymm13[7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm13 = ymm13[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm2[0,1,2],ymm13[3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 768(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpermd %ymm2, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vmovdqa %ymm2, %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[3,0,2,2,7,4,6,6]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm13 = ymm12[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],ymm12[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,3,2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm13[0,1,2],ymm0[3,4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastq {{.*#+}} ymm0 = [25769803777,25769803777,25769803777,25769803777]
+; AVX2-ONLY-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm13 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm15 = ymm10[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],ymm10[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm13 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpalignr $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = mem[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],mem[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm13 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm15 = ymm7[12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10,11],ymm7[28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm13 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpalignr $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = mem[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],mem[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm8, %ymm0, %ymm15
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm12[4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm15 = ymm11[12,13,14,15],ymm9[0,1,2,3,4,5,6,7,8,9,10,11],ymm11[28,29,30,31],ymm9[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm11, %ymm14
+; AVX2-ONLY-NEXT:    vmovdqa %ymm9, %ymm11
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm0, %ymm15
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovdqu (%rsp), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm9[0,1,2,3],ymm8[4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm15 = ymm3[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],ymm3[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpermd %ymm5, %ymm0, %ymm15
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm10[0,1,2,3],ymm7[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm15 = ymm4[12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm15 = ymm13[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm10[0,1],ymm7[0,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3,4],ymm7[5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa {{.*#+}} xmm7 = <4,1,6,u>
+; AVX2-ONLY-NEXT:    vpermd %ymm4, %ymm7, %ymm4
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastq {{.*#+}} ymm6 = [30064771074,30064771074,30064771074,30064771074]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm6, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = mem[0,1],ymm10[0,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0],ymm10[1],ymm4[2,3,4],ymm10[5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm7, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm4 = ymm9[0,1],ymm8[0,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0],ymm8[1],ymm4[2,3,4],ymm8[5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm1, %ymm7, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm5, %ymm6, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = mem[0,1],ymm8[0,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0],ymm8[1],ymm4[2,3,4],ymm8[5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm7, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm14[4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = mem[0,1],ymm12[0,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0],ymm12[1],ymm8[2,3,4],ymm12[5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm7, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = mem[0,1],ymm10[0,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0],ymm10[1],ymm8[2,3,4],ymm10[5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm3, %ymm7, %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm10 = mem[0,1],ymm5[0,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0],ymm5[1],ymm10[2,3,4],ymm5[5],ymm10[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm8, %ymm7, %ymm8
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2],ymm10[3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm9 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vpermd %ymm9, %ymm7, %ymm7
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm9 = mem[0,1],ymm10[0,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2,3,4],ymm10[5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 192(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 128(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 64(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, (%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 224(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 160(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 96(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 32(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 192(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 128(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 64(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, (%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 224(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 160(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 96(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 32(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 128(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 64(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, (%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 192(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 224(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 160(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 96(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 32(%rcx)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm15, (%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 64(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 128(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 192(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 224(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 160(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 96(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 32(%r8)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm6, 224(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm8, 192(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm3, 160(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm2, 128(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm4, 96(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm1, 64(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm13, 32(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm0, (%r9)
+; AVX2-ONLY-NEXT:    addq $2120, %rsp # imm = 0x848
+; AVX2-ONLY-NEXT:    vzeroupper
+; AVX2-ONLY-NEXT:    retq
+;
+; AVX512F-LABEL: load_i32_stride5_vf64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    subq $584, %rsp # imm = 0x248
+; AVX512F-NEXT:    vmovdqa64 1152(%rdi), %zmm11
+; AVX512F-NEXT:    vmovdqa64 1088(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 832(%rdi), %zmm10
+; AVX512F-NEXT:    vmovdqa64 768(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa64 384(%rdi), %zmm15
+; AVX512F-NEXT:    vmovdqa64 512(%rdi), %zmm17
+; AVX512F-NEXT:    vmovdqa64 448(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqa64 (%rdi), %zmm20
+; AVX512F-NEXT:    vmovdqa64 64(%rdi), %zmm13
+; AVX512F-NEXT:    vmovdqa64 128(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqa64 192(%rdi), %zmm21
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [8,13,18,23,28,0,0,3,8,13,18,23,28,0,0,3]
+; AVX512F-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm24
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm4
+; AVX512F-NEXT:    vpermt2d %zmm21, %zmm6, %zmm24
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} ymm19 = <0,5,10,15,20,25,30,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm19, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm6, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm6, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm11, %zmm0, %zmm6
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [9,14,19,24,29,0,0,4,9,14,19,24,29,0,0,4]
+; AVX512F-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm21, %zmm5, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} ymm18 = <17,22,27,0,5,10,15,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm27
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm18, %zmm27
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm5, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm5, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm11, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} ymm26 = <2,7,12,17,22,27,u,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm26, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm22 = [10,15,20,25,30,0,0,5,10,15,20,25,30,0,0,5]
+; AVX512F-NEXT:    # zmm22 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm23
+; AVX512F-NEXT:    vpermt2d %zmm21, %zmm22, %zmm23
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm25
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm22, %zmm25
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm28
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm22, %zmm28
+; AVX512F-NEXT:    vpermi2d %zmm11, %zmm0, %zmm22
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm29 = [27,0,5,10,15,0,17,22,27,0,5,10,15,0,17,22]
+; AVX512F-NEXT:    # zmm29 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm30
+; AVX512F-NEXT:    vpermt2d %zmm4, %zmm29, %zmm30
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [12,17,22,27,0,0,2,7,12,17,22,27,0,0,2,7]
+; AVX512F-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm21, %zmm31, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm17, %zmm21
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm29, %zmm21
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm31, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm17
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm29, %zmm17
+; AVX512F-NEXT:    vpermi2d %zmm0, %zmm11, %zmm29
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm31, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm11 = <3,8,13,18,23,28,u,u>
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm31, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm16
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm11, %zmm16
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} ymm31 = <4,9,14,19,24,29,u,u>
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm31, %zmm20
+; AVX512F-NEXT:    vmovdqa64 320(%rdi), %zmm13
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm19, %zmm7
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm9
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm18, %zmm9
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm12
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm26, %zmm12
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm14
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm11, %zmm14
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm31, %zmm13
+; AVX512F-NEXT:    vmovdqa64 704(%rdi), %zmm15
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm4
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm19, %zmm4
+; AVX512F-NEXT:    vmovdqa64 1024(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 960(%rdi), %zmm1
+; AVX512F-NEXT:    vpermi2d %zmm0, %zmm1, %zmm19
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm18, %zmm5
+; AVX512F-NEXT:    vpermi2d %zmm1, %zmm0, %zmm18
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm26, %zmm8
+; AVX512F-NEXT:    vpermi2d %zmm0, %zmm1, %zmm26
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm10
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm11, %zmm10
+; AVX512F-NEXT:    vpermi2d %zmm0, %zmm1, %zmm11
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm31, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm31, %zmm2
+; AVX512F-NEXT:    movw $8064, %ax # imm = 0x1F80
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqa32 %zmm24, %zmm0 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm3, %zmm7 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm3, %zmm4 {%k1}
+; AVX512F-NEXT:    vmovdqa32 %zmm6, %zmm19 {%k1}
+; AVX512F-NEXT:    vmovdqa64 256(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm15 = [0,1,2,3,4,5,6,7,8,9,10,11,12,17,22,27]
+; AVX512F-NEXT:    vpermt2d %zmm3, %zmm15, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm24
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %zmm31
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm15, %zmm7
+; AVX512F-NEXT:    vmovdqa64 896(%rdi), %zmm6
+; AVX512F-NEXT:    vpermt2d %zmm6, %zmm15, %zmm4
+; AVX512F-NEXT:    vmovdqa64 1216(%rdi), %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm15, %zmm19
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm15, %zmm27 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm15, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm15, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm15, %zmm18 {%k1}
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm15 = [0,1,2,3,4,5,6,7,8,9,10,11,12,18,23,28]
+; AVX512F-NEXT:    vpermt2d %zmm3, %zmm15, %zmm27
+; AVX512F-NEXT:    vpermt2d %zmm6, %zmm15, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm15, %zmm9
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm15, %zmm18
+; AVX512F-NEXT:    movb $7, %al
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm23 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm25 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm28 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm26, %zmm22 {%k1}
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm8 = [0,1,2,3,4,5,6,7,8,9,10,11,12,19,24,29]
+; AVX512F-NEXT:    vpermt2d %zmm3, %zmm8, %zmm23
+; AVX512F-NEXT:    vpermt2d %zmm6, %zmm8, %zmm25
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm8, %zmm28
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm8, %zmm22
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm30 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm17 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm14, %zmm21 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm29 {%k1}
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm8 = [0,1,2,3,4,5,6,7,8,9,10,11,12,20,25,30]
+; AVX512F-NEXT:    vpermt2d %zmm3, %zmm8, %zmm30
+; AVX512F-NEXT:    vpermt2d %zmm6, %zmm8, %zmm17
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm8, %zmm21
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm8, %zmm29
+; AVX512F-NEXT:    movb $56, %al
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm20 {%k1}
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm8 = [0,1,2,3,4,5,6,7,8,9,10,11,16,21,26,31]
+; AVX512F-NEXT:    vpermt2d %zmm3, %zmm8, %zmm20
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm13 {%k1}
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm8, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm1 {%k1}
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm8, %zmm1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
+; AVX512F-NEXT:    vpermt2d %zmm6, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm19, 192(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm4, 128(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm24, (%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm18, 192(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm27, (%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm9, 64(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm5, 128(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm22, 192(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm23, (%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm28, 64(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm29, 192(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm30, (%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm21, 64(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm17, 128(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm2, 128(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm1, 192(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm20, (%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm13, 64(%r9)
+; AVX512F-NEXT:    addq $584, %rsp # imm = 0x248
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: load_i32_stride5_vf64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    subq $584, %rsp # imm = 0x248
+; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %zmm11
+; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm10
+; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm15
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm17
+; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm20
+; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm13
+; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm21
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [8,13,18,23,28,0,0,3,8,13,18,23,28,0,0,3]
+; AVX512BW-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm24
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm4
+; AVX512BW-NEXT:    vpermt2d %zmm21, %zmm6, %zmm24
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} ymm19 = <0,5,10,15,20,25,30,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm19, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm6, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm6, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm11, %zmm0, %zmm6
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [9,14,19,24,29,0,0,4,9,14,19,24,29,0,0,4]
+; AVX512BW-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm21, %zmm5, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} ymm18 = <17,22,27,0,5,10,15,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm27
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm18, %zmm27
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm5, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm5, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm11, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} ymm26 = <2,7,12,17,22,27,u,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm26, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm22 = [10,15,20,25,30,0,0,5,10,15,20,25,30,0,0,5]
+; AVX512BW-NEXT:    # zmm22 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm23
+; AVX512BW-NEXT:    vpermt2d %zmm21, %zmm22, %zmm23
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm25
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm22, %zmm25
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm28
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm22, %zmm28
+; AVX512BW-NEXT:    vpermi2d %zmm11, %zmm0, %zmm22
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm29 = [27,0,5,10,15,0,17,22,27,0,5,10,15,0,17,22]
+; AVX512BW-NEXT:    # zmm29 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm30
+; AVX512BW-NEXT:    vpermt2d %zmm4, %zmm29, %zmm30
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [12,17,22,27,0,0,2,7,12,17,22,27,0,0,2,7]
+; AVX512BW-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm21, %zmm31, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, %zmm21
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm29, %zmm21
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm31, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm17
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm29, %zmm17
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm11, %zmm29
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm31, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm11 = <3,8,13,18,23,28,u,u>
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm31, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm16
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm11, %zmm16
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} ymm31 = <4,9,14,19,24,29,u,u>
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm31, %zmm20
+; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm13
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm19, %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm9
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm18, %zmm9
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm12
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm26, %zmm12
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm14
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm11, %zmm14
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm31, %zmm13
+; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm15
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm4
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm19, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm1
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm19
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm18, %zmm5
+; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm18
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm26, %zmm8
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm26
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm10
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm11, %zmm10
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm11
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm31, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm31, %zmm2
+; AVX512BW-NEXT:    movw $8064, %ax # imm = 0x1F80
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqa32 %zmm24, %zmm0 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm3, %zmm7 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm3, %zmm4 {%k1}
+; AVX512BW-NEXT:    vmovdqa32 %zmm6, %zmm19 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm15 = [0,1,2,3,4,5,6,7,8,9,10,11,12,17,22,27]
+; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm15, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm24
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm31
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm15, %zmm7
+; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm6
+; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm15, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm15, %zmm19
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm15, %zmm27 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm15, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm15, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm15, %zmm18 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm15 = [0,1,2,3,4,5,6,7,8,9,10,11,12,18,23,28]
+; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm15, %zmm27
+; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm15, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm15, %zmm9
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm15, %zmm18
+; AVX512BW-NEXT:    movb $7, %al
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm23 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm25 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm28 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm22 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm8 = [0,1,2,3,4,5,6,7,8,9,10,11,12,19,24,29]
+; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm8, %zmm23
+; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm8, %zmm25
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm8, %zmm28
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm8, %zmm22
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm30 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm17 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, %zmm21 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm29 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm8 = [0,1,2,3,4,5,6,7,8,9,10,11,12,20,25,30]
+; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm8, %zmm30
+; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm8, %zmm17
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm8, %zmm21
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm8, %zmm29
+; AVX512BW-NEXT:    movb $56, %al
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm20 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm8 = [0,1,2,3,4,5,6,7,8,9,10,11,16,21,26,31]
+; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm8, %zmm20
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm13 {%k1}
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm8, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm1 {%k1}
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm8, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
+; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm8, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, 192(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, 128(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, (%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, 192(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, (%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, 64(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, 128(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, 192(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, (%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, 64(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, 192(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, (%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, 64(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, 128(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, 128(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, 192(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, (%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, 64(%r9)
+; AVX512BW-NEXT:    addq $584, %rsp # imm = 0x248
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %wide.vec = load <320 x i32>, ptr %in.vec, align 64
+  %strided.vec0 = shufflevector <320 x i32> %wide.vec, <320 x i32> poison, <64 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75, i32 80, i32 85, i32 90, i32 95, i32 100, i32 105, i32 110, i32 115, i32 120, i32 125, i32 130, i32 135, i32 140, i32 145, i32 150, i32 155, i32 160, i32 165, i32 170, i32 175, i32 180, i32 185, i32 190, i32 195, i32 200, i32 205, i32 210, i32 215, i32 220, i32 225, i32 230, i32 235, i32 240, i32 245, i32 250, i32 255, i32 260, i32 265, i32 270, i32 275, i32 280, i32 285, i32 290, i32 295, i32 300, i32 305, i32 310, i32 315>
+  %strided.vec1 = shufflevector <320 x i32> %wide.vec, <320 x i32> poison, <64 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76, i32 81, i32 86, i32 91, i32 96, i32 101, i32 106, i32 111, i32 116, i32 121, i32 126, i32 131, i32 136, i32 141, i32 146, i32 151, i32 156, i32 161, i32 166, i32 171, i32 176, i32 181, i32 186, i32 191, i32 196, i32 201, i32 206, i32 211, i32 216, i32 221, i32 226, i32 231, i32 236, i32 241, i32 246, i32 251, i32 256, i32 261, i32 266, i32 271, i32 276, i32 281, i32 286, i32 291, i32 296, i32 301, i32 306, i32 311, i32 316>
+  %strided.vec2 = shufflevector <320 x i32> %wide.vec, <320 x i32> poison, <64 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77, i32 82, i32 87, i32 92, i32 97, i32 102, i32 107, i32 112, i32 117, i32 122, i32 127, i32 132, i32 137, i32 142, i32 147, i32 152, i32 157, i32 162, i32 167, i32 172, i32 177, i32 182, i32 187, i32 192, i32 197, i32 202, i32 207, i32 212, i32 217, i32 222, i32 227, i32 232, i32 237, i32 242, i32 247, i32 252, i32 257, i32 262, i32 267, i32 272, i32 277, i32 282, i32 287, i32 292, i32 297, i32 302, i32 307, i32 312, i32 317>
+  %strided.vec3 = shufflevector <320 x i32> %wide.vec, <320 x i32> poison, <64 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78, i32 83, i32 88, i32 93, i32 98, i32 103, i32 108, i32 113, i32 118, i32 123, i32 128, i32 133, i32 138, i32 143, i32 148, i32 153, i32 158, i32 163, i32 168, i32 173, i32 178, i32 183, i32 188, i32 193, i32 198, i32 203, i32 208, i32 213, i32 218, i32 223, i32 228, i32 233, i32 238, i32 243, i32 248, i32 253, i32 258, i32 263, i32 268, i32 273, i32 278, i32 283, i32 288, i32 293, i32 298, i32 303, i32 308, i32 313, i32 318>
+  %strided.vec4 = shufflevector <320 x i32> %wide.vec, <320 x i32> poison, <64 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79, i32 84, i32 89, i32 94, i32 99, i32 104, i32 109, i32 114, i32 119, i32 124, i32 129, i32 134, i32 139, i32 144, i32 149, i32 154, i32 159, i32 164, i32 169, i32 174, i32 179, i32 184, i32 189, i32 194, i32 199, i32 204, i32 209, i32 214, i32 219, i32 224, i32 229, i32 234, i32 239, i32 244, i32 249, i32 254, i32 259, i32 264, i32 269, i32 274, i32 279, i32 284, i32 289, i32 294, i32 299, i32 304, i32 309, i32 314, i32 319>
+  store <64 x i32> %strided.vec0, ptr %out.vec0, align 64
+  store <64 x i32> %strided.vec1, ptr %out.vec1, align 64
+  store <64 x i32> %strided.vec2, ptr %out.vec2, align 64
+  store <64 x i32> %strided.vec3, ptr %out.vec3, align 64
+  store <64 x i32> %strided.vec4, ptr %out.vec4, align 64
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
index e145574..0b7c892 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
@@ -1,5 +1,4 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse2 | FileCheck %s --check-prefixes=SSE,FALLBACK0
 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx  | FileCheck %s --check-prefixes=AVX,AVX1,AVX1-ONLY,FALLBACK1
 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX1,AVX2,AVX2-ONLY,AVX2-SLOW,FALLBACK2
@@ -217,21 +216,18 @@ define void @load_i32_stride6_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <12 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <2 x i32> <i32 0, i32 6>
   %strided.vec1 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <2 x i32> <i32 1, i32 7>
   %strided.vec2 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <2 x i32> <i32 2, i32 8>
   %strided.vec3 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <2 x i32> <i32 3, i32 9>
   %strided.vec4 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <2 x i32> <i32 4, i32 10>
   %strided.vec5 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <2 x i32> <i32 5, i32 11>
-
   store <2 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <2 x i32> %strided.vec3, ptr %out.vec3, align 64
   store <2 x i32> %strided.vec4, ptr %out.vec4, align 64
   store <2 x i32> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -675,21 +671,18 @@ define void @load_i32_stride6_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512DQBW-FAST-NEXT:    vmovdqa %xmm1, (%rax)
 ; AVX512DQBW-FAST-NEXT:    retq
   %wide.vec = load <24 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <4 x i32> <i32 0, i32 6, i32 12, i32 18>
   %strided.vec1 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <4 x i32> <i32 1, i32 7, i32 13, i32 19>
   %strided.vec2 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <4 x i32> <i32 2, i32 8, i32 14, i32 20>
   %strided.vec3 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <4 x i32> <i32 3, i32 9, i32 15, i32 21>
   %strided.vec4 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <4 x i32> <i32 4, i32 10, i32 16, i32 22>
   %strided.vec5 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <4 x i32> <i32 5, i32 11, i32 17, i32 23>
-
   store <4 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <4 x i32> %strided.vec3, ptr %out.vec3, align 64
   store <4 x i32> %strided.vec4, ptr %out.vec4, align 64
   store <4 x i32> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -1197,21 +1190,18 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <48 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <8 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42>
   %strided.vec1 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <8 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43>
   %strided.vec2 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <8 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44>
   %strided.vec3 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <8 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45>
   %strided.vec4 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <8 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46>
   %strided.vec5 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <8 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47>
-
   store <8 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <8 x i32> %strided.vec3, ptr %out.vec3, align 64
   store <8 x i32> %strided.vec4, ptr %out.vec4, align 64
   store <8 x i32> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -2438,21 +2428,18 @@ define void @load_i32_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <96 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <16 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90>
   %strided.vec1 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <16 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91>
   %strided.vec2 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <16 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92>
   %strided.vec3 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <16 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93>
   %strided.vec4 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <16 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94>
   %strided.vec5 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <16 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95>
-
   store <16 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <16 x i32> %strided.vec3, ptr %out.vec3, align 64
   store <16 x i32> %strided.vec4, ptr %out.vec4, align 64
   store <16 x i32> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -5069,21 +5056,5676 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <192 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <32 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90, i32 96, i32 102, i32 108, i32 114, i32 120, i32 126, i32 132, i32 138, i32 144, i32 150, i32 156, i32 162, i32 168, i32 174, i32 180, i32 186>
   %strided.vec1 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <32 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91, i32 97, i32 103, i32 109, i32 115, i32 121, i32 127, i32 133, i32 139, i32 145, i32 151, i32 157, i32 163, i32 169, i32 175, i32 181, i32 187>
   %strided.vec2 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <32 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92, i32 98, i32 104, i32 110, i32 116, i32 122, i32 128, i32 134, i32 140, i32 146, i32 152, i32 158, i32 164, i32 170, i32 176, i32 182, i32 188>
   %strided.vec3 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <32 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93, i32 99, i32 105, i32 111, i32 117, i32 123, i32 129, i32 135, i32 141, i32 147, i32 153, i32 159, i32 165, i32 171, i32 177, i32 183, i32 189>
   %strided.vec4 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <32 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94, i32 100, i32 106, i32 112, i32 118, i32 124, i32 130, i32 136, i32 142, i32 148, i32 154, i32 160, i32 166, i32 172, i32 178, i32 184, i32 190>
   %strided.vec5 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <32 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95, i32 101, i32 107, i32 113, i32 119, i32 125, i32 131, i32 137, i32 143, i32 149, i32 155, i32 161, i32 167, i32 173, i32 179, i32 185, i32 191>
-
   store <32 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i32> %strided.vec2, ptr %out.vec2, align 64
   store <32 x i32> %strided.vec3, ptr %out.vec3, align 64
   store <32 x i32> %strided.vec4, ptr %out.vec4, align 64
   store <32 x i32> %strided.vec5, ptr %out.vec5, align 64
+  ret void
+}
 
+define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
+; SSE-LABEL: load_i32_stride6_vf64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    subq $2184, %rsp # imm = 0x888
+; SSE-NEXT:    movdqa 912(%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 928(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 864(%rdi), %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 880(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 528(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 544(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 480(%rdi), %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 496(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 144(%rdi), %xmm11
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 160(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 96(%rdi), %xmm12
+; SSE-NEXT:    movdqa %xmm12, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa 112(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1248(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1264(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 1296(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1312(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 16(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 64(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 48(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 384(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 400(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 432(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 448(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 768(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 784(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 816(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 832(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1152(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1168(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 1200(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1216(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 288(%rdi), %xmm7
+; SSE-NEXT:    movdqa 304(%rdi), %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 336(%rdi), %xmm10
+; SSE-NEXT:    movdqa 352(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,0,1,1]
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 672(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 688(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 720(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 736(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm3[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
+; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1056(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1072(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 1104(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1120(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm3[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
+; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1440(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1456(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 1488(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1504(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm2[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
+; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 192(%rdi), %xmm3
+; SSE-NEXT:    movdqa 208(%rdi), %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 240(%rdi), %xmm2
+; SSE-NEXT:    movdqa 256(%rdi), %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm2[0,0,1,1]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
+; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 576(%rdi), %xmm6
+; SSE-NEXT:    movdqa 592(%rdi), %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, %xmm8
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    movdqa 624(%rdi), %xmm11
+; SSE-NEXT:    movdqa 640(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm11[0,0,1,1]
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm8[0],xmm14[1]
+; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 960(%rdi), %xmm14
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 976(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm0[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm8[0],xmm14[1],xmm8[1]
+; SSE-NEXT:    movdqa 1008(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1024(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm0[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm1[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm15[0],xmm8[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm14[0],xmm8[1]
+; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1344(%rdi), %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1360(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm0[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
+; SSE-NEXT:    movdqa 1392(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1408(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm1[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm0[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm14[0],xmm8[1],xmm14[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm15[0],xmm8[1]
+; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, (%rsp), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm15[0],xmm8[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm14[0],xmm8[1]
+; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm8
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm15[0],xmm8[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm14[0],xmm8[1]
+; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm7[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm13[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm15[0],xmm10[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm14[0],xmm10[1]
+; SSE-NEXT:    movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm3[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm3, %xmm13
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm4[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm5[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm14[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm8, %xmm7
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm15[0],xmm7[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm14[0],xmm7[1]
+; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm4, %xmm7
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm15[0],xmm7[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm14[0],xmm7[1]
+; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm15[0],xmm3[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm14[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm6[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm12[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm15[0],xmm11[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm14[0],xmm11[1]
+; SSE-NEXT:    movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm15[0],xmm3[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm14[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm14[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm14[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm6[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm12[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
+; SSE-NEXT:    movdqa 80(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm9[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa 32(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa 176(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm9[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    pshufd $238, (%rsp), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa 128(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa 272(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm9[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm13[2,3,2,3]
+; SSE-NEXT:    movdqa 224(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa 368(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm9[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm13[2,3,2,3]
+; SSE-NEXT:    movdqa 320(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
+; SSE-NEXT:    movdqa 464(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm4[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa 416(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
+; SSE-NEXT:    movdqa 560(%rdi), %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm8[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm8[2,3,2,3]
+; SSE-NEXT:    movdqa 512(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa 656(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm4[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa 608(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa 752(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm4[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa 704(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
+; SSE-NEXT:    movdqa 848(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm9[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm11[2,3,2,3]
+; SSE-NEXT:    movdqa 800(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
+; SSE-NEXT:    movdqa 944(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm4[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa 896(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm4[0],xmm14[1],xmm4[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,2,3,3]
+; SSE-NEXT:    movdqa 1040(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm1[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm10[2,3,2,3]
+; SSE-NEXT:    movdqa 992(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm4[0],xmm14[1],xmm4[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa 1136(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm1[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa 1088(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm4[0],xmm14[1],xmm4[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
+; SSE-NEXT:    movdqa 1232(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm1[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm6[2,3,2,3]
+; SSE-NEXT:    movdqa 1184(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm2[0],xmm14[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
+; SSE-NEXT:    movdqa 1328(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm1[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa 1280(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT:    movdqa 1424(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm1[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm15[2,3,2,3]
+; SSE-NEXT:    movdqa 1376(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
+; SSE-NEXT:    movdqa 1520(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm1[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm12[2,3,2,3]
+; SSE-NEXT:    movdqa 1472(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm14[0],xmm5[1],xmm14[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, (%rsp), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm13[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm1[0],xmm12[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm13[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm1[0],xmm6[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
+; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
+; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
+; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
+; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
+; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
+; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
+; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
+; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
+; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
+; SSE-NEXT:    movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
+; SSE-NEXT:    movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
+; SSE-NEXT:    movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
+; SSE-NEXT:    movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r8)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r8)
+; SSE-NEXT:    movapd %xmm2, 240(%r9)
+; SSE-NEXT:    movapd %xmm3, 224(%r9)
+; SSE-NEXT:    movapd %xmm5, 208(%r9)
+; SSE-NEXT:    movapd %xmm6, 192(%r9)
+; SSE-NEXT:    movapd %xmm7, 176(%r9)
+; SSE-NEXT:    movapd %xmm8, 160(%r9)
+; SSE-NEXT:    movapd %xmm9, 144(%r9)
+; SSE-NEXT:    movapd %xmm12, 128(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r9)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movapd %xmm14, 240(%rax)
+; SSE-NEXT:    movapd %xmm13, 224(%rax)
+; SSE-NEXT:    movapd %xmm15, 208(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    addq $2184, %rsp # imm = 0x888
+; SSE-NEXT:    retq
+;
+; AVX1-ONLY-LABEL: load_i32_stride6_vf64:
+; AVX1-ONLY:       # %bb.0:
+; AVX1-ONLY-NEXT:    subq $2568, %rsp # imm = 0xA08
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%rdi), %ymm0, %ymm8
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm8[0,0],ymm1[6,4],ymm8[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm8[2,2],ymm0[6,4],ymm8[6,6]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm6
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm6[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm6[0,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm4[0,0],ymm1[6,4],ymm4[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 672(%rdi), %ymm3, %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[2,0],ymm1[0,0],ymm2[6,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,2],ymm0[6,4],ymm1[6,6]
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm1[0,0],ymm2[6,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1024(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1056(%rdi), %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm2[0,0],ymm1[6,4],ymm2[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[2,2],ymm0[6,4],ymm2[6,6]
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm13[0,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[0,0],ymm1[6,4],ymm2[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1440(%rdi), %ymm0, %ymm11
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm11[0,0],ymm1[6,4],ymm11[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm11[2,2],ymm0[6,4],ymm11[6,6]
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm9
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm9[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm9[0,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1472(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1504(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[0,0],ymm1[6,4],ymm2[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rdi), %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm7[0,0],ymm1[6,4],ymm7[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm7[2,2],ymm0[6,4],ymm7[6,6]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm6[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm6[0,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[0,0],ymm1[6,4],ymm2[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rdi), %ymm0, %ymm5
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm5[0,0],ymm1[6,4],ymm5[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm5[2,2],ymm0[6,4],ymm5[6,6]
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[0,0],ymm1[6,4],ymm2[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 864(%rdi), %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm3[0,0],ymm1[6,4],ymm3[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm3[2,2],ymm0[6,4],ymm3[6,6]
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm1[4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm8[0,0],ymm1[6,4],ymm8[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1248(%rdi), %ymm0, %ymm12
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm12[0,0],ymm1[6,4],ymm12[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm0[2,0],ymm12[2,2],ymm0[6,4],ymm12[6,6]
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm10
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm0[0,1],xmm10[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm10[0,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2],ymm8[3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1280(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1312(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm0[2,0],ymm8[0,0],ymm0[6,4],ymm8[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[3,0],ymm0[1,0],ymm14[7,4],ymm0[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0],ymm0[2,3],ymm14[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[1,0],xmm1[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm1[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2],ymm14[3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm0[3,0],ymm1[1,0],ymm0[7,4],ymm1[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0],ymm1[2,3],ymm14[6,4],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[1,0],xmm1[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm1[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2],ymm14[3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm0[3,0],ymm1[1,0],ymm0[7,4],ymm1[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0],ymm1[2,3],ymm14[6,4],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[1,0],xmm13[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm15[0,2],xmm13[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm14[3,0],ymm11[1,0],ymm14[7,4],ymm11[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm13[2,0],ymm11[2,3],ymm13[6,4],ymm11[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm0[1,0],xmm9[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm13[0,2],xmm9[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm11[3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm11 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm0[3,0],ymm7[1,0],ymm0[7,4],ymm7[5,4]
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, %ymm13
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm9[2,0],ymm7[2,3],ymm9[6,4],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm0[1,0],xmm6[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm9[0,2],xmm6[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm0[3,0],ymm5[1,0],ymm0[7,4],ymm5[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[2,0],ymm5[2,3],ymm6[6,4],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm0[1,0],xmm4[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm6[0,2],xmm4[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm5 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm0[3,0],ymm3[1,0],ymm0[7,4],ymm3[5,4]
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[2,0],ymm3[2,3],ymm4[6,4],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm0[1,0],xmm2[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm4[0,2],xmm2[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,0],ymm12[1,0],ymm3[7,4],ymm12[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm12[2,3],ymm2[6,4],ymm12[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[1,0],xmm10[3,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[0,2],xmm10[1,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm8[1,0],ymm1[7,4],ymm8[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm12
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,0],xmm12[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,0],ymm1[2,0],ymm4[4,4],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm11 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm11[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm11[2,0],ymm1[4,4],ymm11[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm10 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm10[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm10[2,0],ymm1[4,4],ymm10[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm14[2,1],mem[2,0],ymm14[6,5],mem[6,4]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm8[2,0],ymm1[4,4],ymm8[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[2,1],ymm15[2,0],ymm3[6,5],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm7[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm7[2,0],ymm1[4,4],ymm7[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm13[2,1],mem[2,0],ymm13[6,5],mem[6,4]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,0],ymm1[2,0],ymm3[4,4],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm13 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm13[2,0],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm5 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm5[2,0],ymm1[4,4],ymm5[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm6[2,1],ymm9[2,0],ymm6[6,5],ymm9[6,4]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm6
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm4[2,0],xmm6[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm3[0,0],ymm1[2,0],ymm3[4,4],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1],mem[3,1],ymm0[4,5],mem[7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm2[3,1],xmm12[3,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm12 = ymm12[3,1],mem[2,1],ymm12[7,5],mem[6,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm14[0,1,2],ymm12[3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm11[3,1],ymm0[4,5],ymm11[7,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm2[3,1],mem[3,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm12 = ymm14[3,1],mem[2,1],ymm14[7,5],mem[6,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4],ymm11[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[0,1],ymm10[3,1],ymm0[4,5],ymm10[7,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm0[3,1],mem[3,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm12 = ymm10[3,1],mem[2,1],ymm10[7,5],mem[6,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4],ymm11[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm8[3,1],ymm0[4,5],ymm8[7,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = xmm2[3,1],mem[3,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm10 = ymm10[3,1],mem[2,1],ymm10[7,5],mem[6,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm10[2,3,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm10[3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm7[3,1],ymm0[4,5],ymm7[7,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = xmm2[3,1],mem[3,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm2[3,1],ymm15[2,1],ymm2[7,5],ymm15[6,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3,4],ymm7[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,1],ymm1[3,1],ymm3[4,5],ymm1[7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[3,1],xmm6[3,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm9, %ymm7
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,1],ymm9[2,1],ymm3[7,5],ymm9[6,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm5[3,1],ymm0[4,5],ymm5[7,5]
+; AVX1-ONLY-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm13[3,1],mem[3,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm4[3,1],ymm5[2,1],ymm4[7,5],ymm5[6,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1],mem[3,1],ymm0[4,5],mem[7,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[3,1],mem[3,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm6[3,1],ymm8[2,1],ymm6[7,5],ymm8[6,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm8[0],ymm1[2],ymm8[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,0],ymm6[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm9[0,0],ymm2[6,4],ymm9[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[0,2],ymm1[2,0],ymm9[4,6],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 272(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[1],mem[0],ymm1[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,0],ymm2[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps %ymm6, %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm6[0,0],ymm1[6,4],ymm6[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[0,2],ymm1[2,0],ymm6[4,6],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vmovaps 400(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 464(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm5[0],ymm1[2],ymm5[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,0],ymm4[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm5[0,0],ymm2[6,4],ymm5[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[0,2],ymm1[2,0],ymm5[4,6],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm12 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vmovaps 592(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 656(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[1],mem[0],ymm1[2],mem[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm14[0,1],ymm1[2,0],ymm14[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm12[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[2,0],ymm4[0,0],ymm12[6,4],ymm4[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,2],ymm1[2,0],ymm4[4,6],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm11 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vmovaps 784(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 848(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm7[0],ymm1[2],ymm7[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,0],ymm3[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm11[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm11[2,0],ymm3[0,0],ymm11[6,4],ymm3[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm1[2,0],ymm3[4,6],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm10 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vmovaps 976(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 1040(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[1],mem[0],ymm1[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,0],ymm2[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm10[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[2,0],ymm2[0,0],ymm10[6,4],ymm2[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2],ymm1[2,0],ymm2[4,6],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vmovaps 1168(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 1232(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm15[0],ymm1[2],ymm15[2]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,0],ymm7[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[2,0],ymm1[0,0],ymm8[6,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm1[0,2],ymm7[2,0],ymm1[4,6],ymm7[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm7[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vmovaps 1360(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm13[0],xmm0[1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 1424(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm14[1],mem[0],ymm14[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm13[0,1],ymm15[2,0],ymm13[4,5],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm7[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm7[2,0],ymm0[0,0],ymm7[6,4],ymm0[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm0[0,2],ymm13[2,0],ymm0[4,6],ymm13[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1,2,3,4],ymm13[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[3,0],ymm9[1,0],ymm13[7,4],ymm9[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[0,3],ymm13[2,0],ymm9[4,7],ymm13[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm13[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm15[3,1],mem[1,3],ymm15[7,5],mem[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm14[1,1],ymm15[2,0],ymm14[5,5],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[1,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm13[0,1,2,3,4],ymm9[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm9[3,0],ymm6[1,0],ymm9[7,4],ymm6[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,3],ymm13[2,0],ymm6[4,7],ymm13[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm9[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm9[1,1],ymm15[2,0],ymm9[5,5],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[1,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3,4],ymm6[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm9[3,0],ymm5[1,0],ymm9[7,4],ymm5[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,3],ymm13[2,0],ymm5[4,7],ymm13[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm9[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm9[1,1],ymm15[2,0],ymm9[5,5],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[1,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm13[0,1,2,3,4],ymm5[5,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[3,0],ymm4[1,0],ymm12[7,4],ymm4[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,3],ymm12[2,0],ymm4[4,7],ymm12[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm9[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm13 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm9[1,1],ymm13[2,0],ymm9[5,5],ymm13[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm12[1,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3,4],ymm4[5,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[3,0],ymm3[1,0],ymm11[7,4],ymm3[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,3],ymm11[2,0],ymm3[4,7],ymm11[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm9[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm12 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm9[1,1],ymm12[2,0],ymm9[5,5],ymm12[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[1,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm12[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[3,0],ymm2[1,0],ymm10[7,4],ymm2[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,3],ymm10[2,0],ymm2[4,7],ymm10[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm9[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm11 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm9[1,1],ymm11[2,0],ymm9[5,5],ymm11[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[1,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm11[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[3,0],ymm1[1,0],ymm8[7,4],ymm1[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,3],ymm8[2,0],ymm1[4,7],ymm8[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm10 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm9[1,1],ymm10[2,0],ymm9[5,5],ymm10[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[1,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm10[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[3,0],ymm0[1,0],ymm7[7,4],ymm0[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm7[2,0],ymm0[4,7],ymm7[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = xmm7[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm9[1,1],ymm8[2,0],ymm9[5,5],ymm8[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[1,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 192(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 128(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, (%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 224(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 160(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 192(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 128(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, (%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 224(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 160(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, (%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, (%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 64(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 128(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 192(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 224(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 160(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 96(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 32(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 224(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 192(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 160(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 96(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 64(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 32(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, (%r9)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm6, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm14, (%rax)
+; AVX1-ONLY-NEXT:    addq $2568, %rsp # imm = 0xA08
+; AVX1-ONLY-NEXT:    vzeroupper
+; AVX1-ONLY-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: load_i32_stride6_vf64:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    subq $2472, %rsp # imm = 0x9A8
+; AVX2-SLOW-NEXT:    vmovaps 672(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vmovaps 640(%rdi), %ymm4
+; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 320(%rdi), %ymm6
+; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 352(%rdi), %ymm7
+; AVX2-SLOW-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 288(%rdi), %ymm8
+; AVX2-SLOW-NEXT:    vmovaps 256(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 224(%rdi), %ymm0
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 192(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm5 = <0,6,4,u>
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm5, %ymm0
+; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm8[0,1]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,2,4,5,6,6]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm8[2,2,2,2,6,6,6,6]
+; AVX2-SLOW-NEXT:    vmovaps %ymm8, %ymm13
+; AVX2-SLOW-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4],ymm2[5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vbroadcastsd {{.*#+}} ymm15 = [8589934596,8589934596,8589934596,8589934596]
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm4[0,1],ymm3[0,1]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,2,4,5,6,6]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[2,2,2,2,6,6,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 608(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 576(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm5, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 704(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 736(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1056(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1024(%rdi), %ymm0
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,2,4,5,6,6]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 992(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 960(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm5, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 1088(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1120(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1440(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1408(%rdi), %ymm0
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,2,4,5,6,6]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 1376(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1344(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm5, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 1472(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1504(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %ymm0
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm0[0,1],ymm1[0,1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm12[0,1,2,2,4,5,6,6]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovaps (%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm11, %ymm5, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 128(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 160(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm10, %ymm15, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 480(%rdi), %ymm4
+; AVX2-SLOW-NEXT:    vmovaps 448(%rdi), %ymm0
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm0[0,1],ymm4[0,1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm9[0,1,2,2,4,5,6,6]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[2,2,2,2,6,6,6,6]
+; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 416(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 384(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm8, %ymm5, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 512(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 544(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm7, %ymm15, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 864(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vmovaps 832(%rdi), %ymm0
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm0[0,1],ymm3[0,1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[0,1,2,2,4,5,6,6]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[2,2,2,2,6,6,6,6]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 800(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 768(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm5, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 896(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 928(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1248(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1216(%rdi), %ymm0
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,2,4,5,6,6]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 1184(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1152(%rdi), %ymm0
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm2, %ymm5, %ymm5
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm1[3,4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 1280(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1312(%rdi), %ymm5
+; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm15, %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = mem[1,1,2,3,5,5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm13[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0],ymm15[1],ymm0[2,3,4],ymm15[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm0 = <1,7,5,u>
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastsd {{.*#+}} ymm15 = [12884901893,12884901893,12884901893,12884901893]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm13 = mem[1,1,2,3,5,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm13 = mem[1,1,2,3,5,5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm13 = mem[1,1,2,3,5,5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[1,1,2,3,5,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm14[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm11, %ymm0, %ymm11
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm10, %ymm15, %ymm10
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[1,1,2,3,5,5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2,3,4],ymm10[5],ymm9[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm8, %ymm0, %ymm8
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm7, %ymm15, %ymm7
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[1,1,2,3,5,5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3,4],ymm7[5],ymm6[6,7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3,4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm2, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[1,1,2,3,5,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4],ymm2[5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1],ymm5[2,3],mem[4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0,1],ymm7[2,3],ymm15[4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm9[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm13[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1],ymm8[2,3],mem[4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm11[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1],ymm14[2,3],mem[4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = ymm14[0,1],mem[2,3],ymm14[4,5],mem[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $255, (%rsp), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4],ymm15[5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm13[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm11[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vmovaps %ymm14, %ymm13
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2,3,4],ymm8[5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm11[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 272(%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm1 = <4,2,u,u>
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm1, %ymm2
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
+; AVX2-SLOW-NEXT:    # ymm0 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT:    vpermps %ymm14, %ymm0, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 464(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm11[0,1],ymm10[2,3],ymm11[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm0, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm6 = mem[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 656(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = ymm3[0,1],mem[2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm0, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 848(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm0, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $48, (%rsp), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm6 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm6, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1040(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm13 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm13, %ymm0, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps 1232(%rdi), %xmm12
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm10[0,1],ymm12[2,3],ymm10[4,5],ymm12[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm9 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm9, %ymm0, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm11 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm11 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm7 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps 1424(%rdi), %xmm8
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vpermps %ymm11, %ymm1, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm6 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm0, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm4 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 80(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm2[2,3],ymm5[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vpermps %ymm4, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6],mem[7]
+; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm1 = <5,3,u,u>
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
+; AVX2-SLOW-NEXT:    # ymm0 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT:    vpermps %ymm14, %ymm0, %ymm14
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6],mem[7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6],mem[7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6],mem[7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $85, (%rsp), %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6],mem[7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm13, %ymm0, %ymm13
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm12[3],ymm10[4,5,6],ymm12[7]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm9, %ymm0, %ymm9
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT:    vpermps %ymm11, %ymm1, %ymm8
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm0, %ymm6
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3],ymm5[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vpermps %ymm4, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 192(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 128(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 64(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, (%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 224(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 160(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 96(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 32(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 192(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 128(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 64(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, (%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 224(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 160(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 96(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 32(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 128(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 64(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, (%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 192(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 224(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 160(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 96(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 32(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, (%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 64(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 128(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 192(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 224(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 160(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 96(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 32(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 224(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 192(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 160(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 128(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 96(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 64(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 32(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, (%r9)
+; AVX2-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-SLOW-NEXT:    vmovaps %ymm6, 224(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 192(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm13, 160(%rax)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 128(%rax)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 96(%rax)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm1, (%rax)
+; AVX2-SLOW-NEXT:    addq $2472, %rsp # imm = 0x9A8
+; AVX2-SLOW-NEXT:    vzeroupper
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: load_i32_stride6_vf64:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    subq $2472, %rsp # imm = 0x9A8
+; AVX2-FAST-NEXT:    vmovaps 672(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vmovaps 640(%rdi), %ymm4
+; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 320(%rdi), %ymm6
+; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 352(%rdi), %ymm7
+; AVX2-FAST-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 288(%rdi), %ymm8
+; AVX2-FAST-NEXT:    vmovaps 256(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 224(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 192(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm5 = <0,6,4,u>
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm5, %ymm0
+; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm8[0,1]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,2,4,5,6,6]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm8[2,2,2,2,6,6,6,6]
+; AVX2-FAST-NEXT:    vmovaps %ymm8, %ymm13
+; AVX2-FAST-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4],ymm2[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5],ymm7[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vbroadcastsd {{.*#+}} ymm15 = [8589934596,8589934596,8589934596,8589934596]
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm4[0,1],ymm3[0,1]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,2,4,5,6,6]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[2,2,2,2,6,6,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 608(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 576(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm5, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovaps 704(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 736(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1056(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1024(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,2,4,5,6,6]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 992(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 960(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm5, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovaps 1088(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1120(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1440(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1408(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,2,4,5,6,6]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 1376(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1344(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm5, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovaps 1472(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1504(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 64(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm12[0,1,2,2,4,5,6,6]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps (%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm11, %ymm5, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovaps 128(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 160(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm10, %ymm15, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 480(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 448(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm9[0,1,2,2,4,5,6,6]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 416(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 384(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm8, %ymm5, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovaps 512(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 544(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm7, %ymm15, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 864(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vmovaps 832(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm0[0,1],ymm3[0,1]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[0,1,2,2,4,5,6,6]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[2,2,2,2,6,6,6,6]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 800(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 768(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm5, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovaps 896(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 928(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1248(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1216(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,2,4,5,6,6]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 1184(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1152(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm5, %ymm5
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm1[3,4,5],ymm5[6,7]
+; AVX2-FAST-NEXT:    vmovaps 1280(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1312(%rdi), %ymm5
+; AVX2-FAST-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5],ymm5[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm15, %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = mem[1,1,2,3,5,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm13[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0],ymm15[1],ymm0[2,3,4],ymm15[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm0 = <1,7,5,u>
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd {{.*#+}} ymm15 = [12884901893,12884901893,12884901893,12884901893]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm13 = mem[1,1,2,3,5,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm14 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm13 = mem[1,1,2,3,5,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm13 = mem[1,1,2,3,5,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[1,1,2,3,5,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm13 = ymm14[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm11, %ymm0, %ymm11
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5],ymm11[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm10, %ymm15, %ymm10
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[1,1,2,3,5,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm11[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2,3,4],ymm10[5],ymm9[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm8, %ymm0, %ymm8
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5],ymm8[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm7, %ymm15, %ymm7
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[1,1,2,3,5,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3,4],ymm7[5],ymm6[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm0, %ymm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3,4,5],ymm4[6,7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[1,1,2,3,5,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4],ymm2[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm0 = [2,0,6,4,2,0,6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,1],ymm5[2,3],mem[4,5],ymm5[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm5[2,3],ymm9[4,5],ymm5[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm8[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm12[2,3],ymm2[4,5],ymm12[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm15[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $238, (%rsp), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,1],ymm14[2,3],mem[4,5],ymm14[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,1],ymm11[2,3],mem[4,5],ymm11[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm14[0,1],ymm11[2,3],ymm14[4,5],ymm11[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm6[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm13[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1],ymm1[2,3,4],ymm9[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $255, (%rsp), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovaps %ymm11, %ymm12
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm11[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vmovaps %ymm14, %ymm13
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2,3,4],ymm8[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm11[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 272(%rdi), %xmm0
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm1 = <4,2,u,u>
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm1, %ymm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
+; AVX2-FAST-NEXT:    # ymm0 = mem[0,1,0,1]
+; AVX2-FAST-NEXT:    vpermps %ymm14, %ymm0, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 464(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm11[0,1],ymm10[2,3],ymm11[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm0, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm6 = mem[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 656(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = ymm3[0,1],mem[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm0, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 848(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm0, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm6 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1040(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm13 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm13, %ymm0, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps $48, (%rsp), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = ymm15[0,1,2,3],mem[4,5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm10 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovaps 1232(%rdi), %xmm12
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm10[0,1],ymm12[2,3],ymm10[4,5],ymm12[6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm9 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm0, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm11 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm7 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovaps 1424(%rdi), %xmm8
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5],ymm8[6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vpermps %ymm11, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm6 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm0, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm4 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovaps 80(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm2[2,3],ymm5[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm1, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6],mem[7]
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm1 = <5,3,u,u>
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
+; AVX2-FAST-NEXT:    # ymm0 = mem[0,1,0,1]
+; AVX2-FAST-NEXT:    vpermps %ymm14, %ymm0, %ymm14
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6],mem[7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6],mem[7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6],mem[7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6],mem[7]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm13, %ymm0, %ymm13
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm12[3],ymm10[4,5,6],ymm12[7]
+; AVX2-FAST-NEXT:    vpermps (%rsp), %ymm1, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm0, %ymm9
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5,6],ymm8[7]
+; AVX2-FAST-NEXT:    vpermps %ymm11, %ymm1, %ymm8
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm0, %ymm6
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3],ymm5[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm1, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 192(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 128(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 64(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, (%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 224(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 160(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 96(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 32(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 192(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 128(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 64(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, (%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 224(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 160(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 96(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 32(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 128(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 64(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, (%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 192(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 224(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 160(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 96(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 32(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, (%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 64(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 128(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 192(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 224(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 160(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 96(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 32(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 224(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 192(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 160(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 128(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 96(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 64(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 32(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, (%r9)
+; AVX2-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-FAST-NEXT:    vmovaps %ymm6, 224(%rax)
+; AVX2-FAST-NEXT:    vmovaps %ymm9, 192(%rax)
+; AVX2-FAST-NEXT:    vmovaps %ymm13, 160(%rax)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 128(%rax)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 96(%rax)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX2-FAST-NEXT:    vmovaps %ymm1, (%rax)
+; AVX2-FAST-NEXT:    addq $2472, %rsp # imm = 0x9A8
+; AVX2-FAST-NEXT:    vzeroupper
+; AVX2-FAST-NEXT:    retq
+;
+; AVX2-FAST-PERLANE-LABEL: load_i32_stride6_vf64:
+; AVX2-FAST-PERLANE:       # %bb.0:
+; AVX2-FAST-PERLANE-NEXT:    subq $2472, %rsp # imm = 0x9A8
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 672(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 640(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 320(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 352(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 288(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 256(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm5 = <0,6,4,u>
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm0, %ymm5, %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm8[0,1]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,2,4,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm8[2,2,2,2,6,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4],ymm2[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{.*#+}} ymm15 = [8589934596,8589934596,8589934596,8589934596]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm4[0,1],ymm3[0,1]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,2,4,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[2,2,2,2,6,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 608(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 576(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 704(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 736(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1056(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1024(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,2,4,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 992(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 960(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1088(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1120(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1440(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1408(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,2,4,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1376(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1344(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1472(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1504(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm12[0,1,2,2,4,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm11, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm10, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 480(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 448(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm0[0,1],ymm4[0,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm9[0,1,2,2,4,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[2,2,2,2,6,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 416(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 384(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm8, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 512(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 544(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm7, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 864(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 832(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm0[0,1],ymm3[0,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[0,1,2,2,4,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[2,2,2,2,6,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 800(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 768(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 896(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 928(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1248(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1216(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,2,4,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1184(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1152(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm2, %ymm5, %ymm5
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm1[3,4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1280(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1312(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm15, %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[1,1,2,3,5,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm13[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0],ymm15[1],ymm0[2,3,4],ymm15[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm0 = <1,7,5,u>
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{.*#+}} ymm15 = [12884901893,12884901893,12884901893,12884901893]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm13 = mem[1,1,2,3,5,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm13 = mem[1,1,2,3,5,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm13 = mem[1,1,2,3,5,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[1,1,2,3,5,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm14[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm11, %ymm0, %ymm11
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm10, %ymm15, %ymm10
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[1,1,2,3,5,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2,3,4],ymm10[5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm8, %ymm0, %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm7, %ymm15, %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[1,1,2,3,5,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3,4],ymm7[5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm2, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $229, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[1,1,2,3,5,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4],ymm2[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1],ymm5[2,3],mem[4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0,1],ymm7[2,3],ymm15[4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm9[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm13[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1],ymm8[2,3],mem[4,5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm11[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1],ymm14[2,3],mem[4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm14[0,1],mem[2,3],ymm14[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $255, (%rsp), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4],ymm15[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm13[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm11[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm14, %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2,3,4],ymm8[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm11[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 272(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm1 = <4,2,u,u>
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm1, %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm14, %ymm0, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 464(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm11[0,1],ymm10[2,3],ymm11[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm0, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm6 = mem[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 656(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = ymm3[0,1],mem[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm0, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 848(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm0, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $48, (%rsp), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm6 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1040(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm13 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm13, %ymm0, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1232(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm10[0,1],ymm12[2,3],ymm10[4,5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm1, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm9 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm9, %ymm0, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm11 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm7 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1424(%rdi), %xmm8
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm11, %ymm1, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm6 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm0, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm4 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 80(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm2[2,3],ymm5[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm4, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6],mem[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm1 = <5,3,u,u>
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm14, %ymm0, %ymm14
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6],mem[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6],mem[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6],mem[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $85, (%rsp), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $136, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6],mem[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm13, %ymm0, %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm12[3],ymm10[4,5,6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm9, %ymm0, %ymm9
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm11, %ymm1, %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm0, %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3],ymm5[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm4, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 192(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 128(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 64(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, (%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 224(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 160(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 96(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 32(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 192(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 128(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 64(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, (%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 224(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 160(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 96(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 32(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 128(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 64(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, (%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 192(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 224(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 160(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 96(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, (%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 64(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 128(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 192(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 224(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 160(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 96(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 32(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 224(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 192(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 160(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 128(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 96(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 64(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 32(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, (%r9)
+; AVX2-FAST-PERLANE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm6, 224(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 192(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm13, 160(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 128(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 96(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, (%rax)
+; AVX2-FAST-PERLANE-NEXT:    addq $2472, %rsp # imm = 0x9A8
+; AVX2-FAST-PERLANE-NEXT:    vzeroupper
+; AVX2-FAST-PERLANE-NEXT:    retq
+;
+; AVX512F-LABEL: load_i32_stride6_vf64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    subq $2632, %rsp # imm = 0xA48
+; AVX512F-NEXT:    vmovdqa64 1472(%rdi), %zmm13
+; AVX512F-NEXT:    vmovdqa64 1408(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa64 1088(%rdi), %zmm11
+; AVX512F-NEXT:    vmovdqa64 1024(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 512(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqa64 704(%rdi), %zmm20
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqa64 320(%rdi), %zmm18
+; AVX512F-NEXT:    vmovdqa64 256(%rdi), %zmm5
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
+; AVX512F-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
+; AVX512F-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm7, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm7, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm3, %zmm6, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm7, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm13, %zmm1, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
+; AVX512F-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm8, %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
+; AVX512F-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm8, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm8, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm13, %zmm1, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm7, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm4, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm5, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm1, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm7, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm4, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm5, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm1, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
+; AVX512F-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm8, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm0, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm4
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm8, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm4
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm8, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm13, %zmm1, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm3, %zmm9, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm0, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm18 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
+; AVX512F-NEXT:    # zmm18 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm18, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
+; AVX512F-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm31, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
+; AVX512F-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm4, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
+; AVX512F-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm1, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 128(%rdi), %zmm24
+; AVX512F-NEXT:    vmovdqa64 192(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm24, %zmm6, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm24, %zmm9, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm28
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm18, %zmm28
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm29
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm31, %zmm29
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm4, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm1, %zmm24
+; AVX512F-NEXT:    vmovdqa64 896(%rdi), %zmm20
+; AVX512F-NEXT:    vmovdqa64 960(%rdi), %zmm10
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm6, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1280(%rdi), %zmm17
+; AVX512F-NEXT:    vmovdqa64 1344(%rdi), %zmm0
+; AVX512F-NEXT:    vpermi2d %zmm17, %zmm0, %zmm6
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm9, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm17, %zmm0, %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm26
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm18, %zmm26
+; AVX512F-NEXT:    vpermi2d %zmm0, %zmm17, %zmm18
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm27
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm31, %zmm27
+; AVX512F-NEXT:    vpermi2d %zmm0, %zmm17, %zmm31
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm0, %zmm17, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm1, %zmm17
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm1, %zmm20
+; AVX512F-NEXT:    vmovdqa64 448(%rdi), %zmm11
+; AVX512F-NEXT:    vmovdqa64 384(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm10 = <0,6,12,18,24,30,u,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm19
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm10, %zmm19
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = <1,7,13,19,25,31,u,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm21
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm3, %zmm21
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm12 = <2,8,14,20,26,u,u,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm12, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm15 = <3,9,15,21,27,u,u,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm15, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = <20,26,0,6,12,u,u,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm22
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm4, %zmm22
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = <21,27,1,7,13,u,u,u>
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm2, %zmm11
+; AVX512F-NEXT:    vmovdqa64 (%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa64 64(%rdi), %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm14
+; AVX512F-NEXT:    vpermt2d %zmm8, %zmm10, %zmm14
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm13
+; AVX512F-NEXT:    vpermt2d %zmm8, %zmm3, %zmm13
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm8, %zmm12, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm8, %zmm15, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm16
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm4, %zmm16
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm2, %zmm8
+; AVX512F-NEXT:    vmovdqa64 832(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqa64 768(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm6
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm10, %zmm6
+; AVX512F-NEXT:    vmovdqa64 1216(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa64 1152(%rdi), %zmm0
+; AVX512F-NEXT:    vpermi2d %zmm1, %zmm0, %zmm10
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm3, %zmm7
+; AVX512F-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm23
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm12, %zmm23
+; AVX512F-NEXT:    vpermi2d %zmm1, %zmm0, %zmm12
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm25
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm15, %zmm25
+; AVX512F-NEXT:    vpermi2d %zmm1, %zmm0, %zmm15
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm9
+; AVX512F-NEXT:    vpermt2d %zmm5, %zmm4, %zmm9
+; AVX512F-NEXT:    vpermi2d %zmm0, %zmm1, %zmm4
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} ymm30 = <21,27,1,7,13,u,u,u>
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm30, %zmm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm0 = <21,27,1,7,13,u,u,u>
+; AVX512F-NEXT:    vpermt2d %zmm5, %zmm0, %zmm2
+; AVX512F-NEXT:    movb $56, %al
+; AVX512F-NEXT:    kmovw %eax, %k2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm14 {%k2}
+; AVX512F-NEXT:    movw $-2048, %ax # imm = 0xF800
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm14 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm19 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm19 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm6 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm6 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm10 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm10 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm7 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm13 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm13 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm3 {%k1}
+; AVX512F-NEXT:    movw $31, %ax
+; AVX512F-NEXT:    kmovw %eax, %k2
+; AVX512F-NEXT:    vmovdqa32 %zmm23, %zmm26 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm26 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm23
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm28 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm28 {%k1}
+; AVX512F-NEXT:    vmovdqa32 %zmm12, %zmm18 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm18 {%k1}
+; AVX512F-NEXT:    vmovdqa32 %zmm25, %zmm27 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm27 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm29 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm29 {%k1}
+; AVX512F-NEXT:    vmovdqa32 %zmm15, %zmm31 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm31 {%k1}
+; AVX512F-NEXT:    movw $992, %ax # imm = 0x3E0
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    movb $-32, %al
+; AVX512F-NEXT:    kmovw %eax, %k2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm22 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm22 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm11 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k2}
+; AVX512F-NEXT:    vmovdqa32 %zmm24, %zmm8 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k2}
+; AVX512F-NEXT:    vmovdqa32 %zmm17, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512F-NEXT:    vmovdqa32 %zmm20, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k2}
+; AVX512F-NEXT:    vmovdqa64 %zmm10, 192(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm6, 128(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm19, 64(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm14, (%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm3, 192(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm13, (%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm21, 64(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm7, 128(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm18, 192(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm28, (%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm23, 64(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm26, 128(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm31, 192(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm29, (%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm5, 64(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm27, 128(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm4, 192(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm16, (%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm22, 64(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm9, 128(%r9)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovdqa64 %zmm2, 128(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm1, 192(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm8, (%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm11, 64(%rax)
+; AVX512F-NEXT:    addq $2632, %rsp # imm = 0xA48
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: load_i32_stride6_vf64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    subq $2632, %rsp # imm = 0xA48
+; AVX512BW-NEXT:    vmovdqa64 1472(%rdi), %zmm13
+; AVX512BW-NEXT:    vmovdqa64 1408(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %zmm11
+; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm20
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm18
+; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm5
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
+; AVX512BW-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
+; AVX512BW-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm7, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm7, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm6, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm7, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm13, %zmm1, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
+; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm8, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
+; AVX512BW-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm8, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm8, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm13, %zmm1, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm7, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm4, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm5, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm7, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm4, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm5, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
+; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm8, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm4
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm8, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm4
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm8, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm13, %zmm1, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm9, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm0, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm18 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
+; AVX512BW-NEXT:    # zmm18 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm18, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
+; AVX512BW-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm31, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
+; AVX512BW-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm4, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
+; AVX512BW-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm1, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm24
+; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm24, %zmm6, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm24, %zmm9, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm28
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm18, %zmm28
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm29
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm31, %zmm29
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm4, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm1, %zmm24
+; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm20
+; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm10
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm6, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1280(%rdi), %zmm17
+; AVX512BW-NEXT:    vmovdqa64 1344(%rdi), %zmm0
+; AVX512BW-NEXT:    vpermi2d %zmm17, %zmm0, %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm9, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm17, %zmm0, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm26
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm18, %zmm26
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm17, %zmm18
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm27
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm31, %zmm27
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm17, %zmm31
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm4, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm17, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm1, %zmm17
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm1, %zmm20
+; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm11
+; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm10 = <0,6,12,18,24,30,u,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm19
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm10, %zmm19
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm3 = <1,7,13,19,25,31,u,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm21
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm3, %zmm21
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm12 = <2,8,14,20,26,u,u,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm12, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm15 = <3,9,15,21,27,u,u,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm15, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm4 = <20,26,0,6,12,u,u,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm22
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm4, %zmm22
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = <21,27,1,7,13,u,u,u>
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm2, %zmm11
+; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm14
+; AVX512BW-NEXT:    vpermt2d %zmm8, %zmm10, %zmm14
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm13
+; AVX512BW-NEXT:    vpermt2d %zmm8, %zmm3, %zmm13
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm8, %zmm12, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm8, %zmm15, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm16
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm4, %zmm16
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm2, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm6
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm10, %zmm6
+; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %zmm0
+; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm10
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm3, %zmm7
+; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm23
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm12, %zmm23
+; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm12
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm25
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm15, %zmm25
+; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm15
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm9
+; AVX512BW-NEXT:    vpermt2d %zmm5, %zmm4, %zmm9
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} ymm30 = <21,27,1,7,13,u,u,u>
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm30, %zmm1
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm0 = <21,27,1,7,13,u,u,u>
+; AVX512BW-NEXT:    vpermt2d %zmm5, %zmm0, %zmm2
+; AVX512BW-NEXT:    movb $56, %al
+; AVX512BW-NEXT:    kmovd %eax, %k2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm14 {%k2}
+; AVX512BW-NEXT:    movw $-2048, %ax # imm = 0xF800
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm14 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm19 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm19 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm6 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm6 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm10 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm10 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm7 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm13 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm13 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm3 {%k1}
+; AVX512BW-NEXT:    movw $31, %ax
+; AVX512BW-NEXT:    kmovd %eax, %k2
+; AVX512BW-NEXT:    vmovdqa32 %zmm23, %zmm26 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm26 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm23
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm28 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm28 {%k1}
+; AVX512BW-NEXT:    vmovdqa32 %zmm12, %zmm18 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm18 {%k1}
+; AVX512BW-NEXT:    vmovdqa32 %zmm25, %zmm27 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm27 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm29 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm29 {%k1}
+; AVX512BW-NEXT:    vmovdqa32 %zmm15, %zmm31 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm31 {%k1}
+; AVX512BW-NEXT:    movw $992, %ax # imm = 0x3E0
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    movb $-32, %al
+; AVX512BW-NEXT:    kmovd %eax, %k2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm22 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm22 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm11 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k2}
+; AVX512BW-NEXT:    vmovdqa32 %zmm24, %zmm8 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k2}
+; AVX512BW-NEXT:    vmovdqa32 %zmm17, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512BW-NEXT:    vmovdqa32 %zmm20, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, 192(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, 128(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, 64(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, (%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, 192(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, (%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, 64(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, 128(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, 192(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, (%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, 64(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, 128(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, 192(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, (%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, 64(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, 128(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, 192(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, (%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, 64(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, 128(%r9)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, 128(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, 192(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, (%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, 64(%rax)
+; AVX512BW-NEXT:    addq $2632, %rsp # imm = 0xA48
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %wide.vec = load <384 x i32>, ptr %in.vec, align 64
+  %strided.vec0 = shufflevector <384 x i32> %wide.vec, <384 x i32> poison, <64 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90, i32 96, i32 102, i32 108, i32 114, i32 120, i32 126, i32 132, i32 138, i32 144, i32 150, i32 156, i32 162, i32 168, i32 174, i32 180, i32 186, i32 192, i32 198, i32 204, i32 210, i32 216, i32 222, i32 228, i32 234, i32 240, i32 246, i32 252, i32 258, i32 264, i32 270, i32 276, i32 282, i32 288, i32 294, i32 300, i32 306, i32 312, i32 318, i32 324, i32 330, i32 336, i32 342, i32 348, i32 354, i32 360, i32 366, i32 372, i32 378>
+  %strided.vec1 = shufflevector <384 x i32> %wide.vec, <384 x i32> poison, <64 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91, i32 97, i32 103, i32 109, i32 115, i32 121, i32 127, i32 133, i32 139, i32 145, i32 151, i32 157, i32 163, i32 169, i32 175, i32 181, i32 187, i32 193, i32 199, i32 205, i32 211, i32 217, i32 223, i32 229, i32 235, i32 241, i32 247, i32 253, i32 259, i32 265, i32 271, i32 277, i32 283, i32 289, i32 295, i32 301, i32 307, i32 313, i32 319, i32 325, i32 331, i32 337, i32 343, i32 349, i32 355, i32 361, i32 367, i32 373, i32 379>
+  %strided.vec2 = shufflevector <384 x i32> %wide.vec, <384 x i32> poison, <64 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92, i32 98, i32 104, i32 110, i32 116, i32 122, i32 128, i32 134, i32 140, i32 146, i32 152, i32 158, i32 164, i32 170, i32 176, i32 182, i32 188, i32 194, i32 200, i32 206, i32 212, i32 218, i32 224, i32 230, i32 236, i32 242, i32 248, i32 254, i32 260, i32 266, i32 272, i32 278, i32 284, i32 290, i32 296, i32 302, i32 308, i32 314, i32 320, i32 326, i32 332, i32 338, i32 344, i32 350, i32 356, i32 362, i32 368, i32 374, i32 380>
+  %strided.vec3 = shufflevector <384 x i32> %wide.vec, <384 x i32> poison, <64 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93, i32 99, i32 105, i32 111, i32 117, i32 123, i32 129, i32 135, i32 141, i32 147, i32 153, i32 159, i32 165, i32 171, i32 177, i32 183, i32 189, i32 195, i32 201, i32 207, i32 213, i32 219, i32 225, i32 231, i32 237, i32 243, i32 249, i32 255, i32 261, i32 267, i32 273, i32 279, i32 285, i32 291, i32 297, i32 303, i32 309, i32 315, i32 321, i32 327, i32 333, i32 339, i32 345, i32 351, i32 357, i32 363, i32 369, i32 375, i32 381>
+  %strided.vec4 = shufflevector <384 x i32> %wide.vec, <384 x i32> poison, <64 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94, i32 100, i32 106, i32 112, i32 118, i32 124, i32 130, i32 136, i32 142, i32 148, i32 154, i32 160, i32 166, i32 172, i32 178, i32 184, i32 190, i32 196, i32 202, i32 208, i32 214, i32 220, i32 226, i32 232, i32 238, i32 244, i32 250, i32 256, i32 262, i32 268, i32 274, i32 280, i32 286, i32 292, i32 298, i32 304, i32 310, i32 316, i32 322, i32 328, i32 334, i32 340, i32 346, i32 352, i32 358, i32 364, i32 370, i32 376, i32 382>
+  %strided.vec5 = shufflevector <384 x i32> %wide.vec, <384 x i32> poison, <64 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95, i32 101, i32 107, i32 113, i32 119, i32 125, i32 131, i32 137, i32 143, i32 149, i32 155, i32 161, i32 167, i32 173, i32 179, i32 185, i32 191, i32 197, i32 203, i32 209, i32 215, i32 221, i32 227, i32 233, i32 239, i32 245, i32 251, i32 257, i32 263, i32 269, i32 275, i32 281, i32 287, i32 293, i32 299, i32 305, i32 311, i32 317, i32 323, i32 329, i32 335, i32 341, i32 347, i32 353, i32 359, i32 365, i32 371, i32 377, i32 383>
+  store <64 x i32> %strided.vec0, ptr %out.vec0, align 64
+  store <64 x i32> %strided.vec1, ptr %out.vec1, align 64
+  store <64 x i32> %strided.vec2, ptr %out.vec2, align 64
+  store <64 x i32> %strided.vec3, ptr %out.vec3, align 64
+  store <64 x i32> %strided.vec4, ptr %out.vec4, align 64
+  store <64 x i32> %strided.vec5, ptr %out.vec5, align 64
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
index fc2f628..7e7a79e 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
@@ -255,7 +255,6 @@ define void @load_i32_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <14 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> <i32 0, i32 7>
   %strided.vec1 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> <i32 1, i32 8>
   %strided.vec2 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> <i32 2, i32 9>
@@ -263,7 +262,6 @@ define void @load_i32_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> <i32 4, i32 11>
   %strided.vec5 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> <i32 5, i32 12>
   %strided.vec6 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> <i32 6, i32 13>
-
   store <2 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i32> %strided.vec2, ptr %out.vec2, align 64
@@ -271,7 +269,6 @@ define void @load_i32_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <2 x i32> %strided.vec4, ptr %out.vec4, align 64
   store <2 x i32> %strided.vec5, ptr %out.vec5, align 64
   store <2 x i32> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -1023,7 +1020,6 @@ define void @load_i32_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512DQBW-FAST-NEXT:    vmovdqa %xmm0, (%rax)
 ; AVX512DQBW-FAST-NEXT:    retq
   %wide.vec = load <28 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> <i32 0, i32 7, i32 14, i32 21>
   %strided.vec1 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> <i32 1, i32 8, i32 15, i32 22>
   %strided.vec2 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> <i32 2, i32 9, i32 16, i32 23>
@@ -1031,7 +1027,6 @@ define void @load_i32_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> <i32 4, i32 11, i32 18, i32 25>
   %strided.vec5 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> <i32 5, i32 12, i32 19, i32 26>
   %strided.vec6 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> <i32 6, i32 13, i32 20, i32 27>
-
   store <4 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i32> %strided.vec2, ptr %out.vec2, align 64
@@ -1039,7 +1034,6 @@ define void @load_i32_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <4 x i32> %strided.vec4, ptr %out.vec4, align 64
   store <4 x i32> %strided.vec5, ptr %out.vec5, align 64
   store <4 x i32> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -1645,7 +1639,6 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <56 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49>
   %strided.vec1 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50>
   %strided.vec2 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51>
@@ -1653,7 +1646,6 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53>
   %strided.vec5 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54>
   %strided.vec6 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55>
-
   store <8 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i32> %strided.vec2, ptr %out.vec2, align 64
@@ -1661,7 +1653,6 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <8 x i32> %strided.vec4, ptr %out.vec4, align 64
   store <8 x i32> %strided.vec5, ptr %out.vec5, align 64
   store <8 x i32> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -3147,7 +3138,6 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <112 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49, i32 56, i32 63, i32 70, i32 77, i32 84, i32 91, i32 98, i32 105>
   %strided.vec1 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50, i32 57, i32 64, i32 71, i32 78, i32 85, i32 92, i32 99, i32 106>
   %strided.vec2 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51, i32 58, i32 65, i32 72, i32 79, i32 86, i32 93, i32 100, i32 107>
@@ -3155,7 +3145,6 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   %strided.vec4 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53, i32 60, i32 67, i32 74, i32 81, i32 88, i32 95, i32 102, i32 109>
   %strided.vec5 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54, i32 61, i32 68, i32 75, i32 82, i32 89, i32 96, i32 103, i32 110>
   %strided.vec6 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55, i32 62, i32 69, i32 76, i32 83, i32 90, i32 97, i32 104, i32 111>
-
   store <16 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i32> %strided.vec2, ptr %out.vec2, align 64
@@ -3163,7 +3152,6 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   store <16 x i32> %strided.vec4, ptr %out.vec4, align 64
   store <16 x i32> %strided.vec5, ptr %out.vec5, align 64
   store <16 x i32> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -6384,7 +6372,6 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <224 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49, i32 56, i32 63, i32 70, i32 77, i32 84, i32 91, i32 98, i32 105, i32 112, i32 119, i32 126, i32 133, i32 140, i32 147, i32 154, i32 161, i32 168, i32 175, i32 182, i32 189, i32 196, i32 203, i32 210, i32 217>
   %strided.vec1 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50, i32 57, i32 64, i32 71, i32 78, i32 85, i32 92, i32 99, i32 106, i32 113, i32 120, i32 127, i32 134, i32 141, i32 148, i32 155, i32 162, i32 169, i32 176, i32 183, i32 190, i32 197, i32 204, i32 211, i32 218>
   %strided.vec2 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51, i32 58, i32 65, i32 72, i32 79, i32 86, i32 93, i32 100, i32 107, i32 114, i32 121, i32 128, i32 135, i32 142, i32 149, i32 156, i32 163, i32 170, i32 177, i32 184, i32 191, i32 198, i32 205, i32 212, i32 219>
@@ -6392,7 +6379,6 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   %strided.vec4 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53, i32 60, i32 67, i32 74, i32 81, i32 88, i32 95, i32 102, i32 109, i32 116, i32 123, i32 130, i32 137, i32 144, i32 151, i32 158, i32 165, i32 172, i32 179, i32 186, i32 193, i32 200, i32 207, i32 214, i32 221>
   %strided.vec5 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54, i32 61, i32 68, i32 75, i32 82, i32 89, i32 96, i32 103, i32 110, i32 117, i32 124, i32 131, i32 138, i32 145, i32 152, i32 159, i32 166, i32 173, i32 180, i32 187, i32 194, i32 201, i32 208, i32 215, i32 222>
   %strided.vec6 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55, i32 62, i32 69, i32 76, i32 83, i32 90, i32 97, i32 104, i32 111, i32 118, i32 125, i32 132, i32 139, i32 146, i32 153, i32 160, i32 167, i32 174, i32 181, i32 188, i32 195, i32 202, i32 209, i32 216, i32 223>
-
   store <32 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i32> %strided.vec2, ptr %out.vec2, align 64
@@ -6400,7 +6386,6740 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   store <32 x i32> %strided.vec4, ptr %out.vec4, align 64
   store <32 x i32> %strided.vec5, ptr %out.vec5, align 64
   store <32 x i32> %strided.vec6, ptr %out.vec6, align 64
+  ret void
+}
 
+define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind {
+; SSE-LABEL: load_i32_stride7_vf64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    subq $2456, %rsp # imm = 0x998
+; SSE-NEXT:    movdqa 1088(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1056(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1008(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1024(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 640(%rdi), %xmm12
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 608(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 560(%rdi), %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 576(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 192(%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 160(%rdi), %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 112(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 128(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1456(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1472(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movdqa 1536(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1504(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 16(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa 80(%rdi), %xmm14
+; SSE-NEXT:    movdqa 48(%rdi), %xmm15
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 448(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa 464(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa 528(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 496(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 896(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 912(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa 976(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 944(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1344(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1360(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa 1424(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1392(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 336(%rdi), %xmm11
+; SSE-NEXT:    movdqa 352(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa 416(%rdi), %xmm8
+; SSE-NEXT:    movdqa 384(%rdi), %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 784(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 800(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa 864(%rdi), %xmm13
+; SSE-NEXT:    movdqa 832(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1232(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1248(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa 1312(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1280(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1680(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1696(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa 1760(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1728(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 224(%rdi), %xmm7
+; SSE-NEXT:    movdqa 240(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa 304(%rdi), %xmm3
+; SSE-NEXT:    movdqa 272(%rdi), %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 672(%rdi), %xmm10
+; SSE-NEXT:    movdqa 688(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT:    movdqa 752(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 720(%rdi), %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm4[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1120(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1136(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT:    movdqa 1200(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1168(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm5[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1568(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1584(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT:    movdqa 1648(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1616(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm5[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm12[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa 144(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm14[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm4[2],xmm15[3],xmm4[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[1,1,1,1]
+; SSE-NEXT:    movdqa 32(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm4[0],xmm15[1]
+; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm4[2],xmm9[3],xmm4[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[1,1,1,1]
+; SSE-NEXT:    movdqa 368(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm4[0],xmm9[1]
+; SSE-NEXT:    movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm4[2],xmm6[3],xmm4[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[1,1,1,1]
+; SSE-NEXT:    movdqa 256(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
+; SSE-NEXT:    movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa 592(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE-NEXT:    movdqa (%rsp), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[1,1,1,1]
+; SSE-NEXT:    movdqa 480(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa 816(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm10[1,1,1,1]
+; SSE-NEXT:    movdqa 704(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm9[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa 1040(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm9[2],xmm1[3],xmm9[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa 928(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm3[0],xmm9[1],xmm3[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm9[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm1[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm9[2],xmm1[3],xmm9[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa 1264(%rdi), %xmm15
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm15[0],xmm9[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm9[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm13[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm9[2],xmm1[3],xmm9[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm14[1,1,1,1]
+; SSE-NEXT:    movdqa 1152(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm3[0],xmm9[1],xmm3[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm9[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm9[2],xmm1[3],xmm9[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa 1488(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm3[0],xmm9[1],xmm3[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm9[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm6[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm9[2],xmm3[3],xmm9[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa 1376(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm9[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm1[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm9[2],xmm3[3],xmm9[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa 1712(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm9[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm8[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm9[2],xmm3[3],xmm9[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm1[1,1,1,1]
+; SSE-NEXT:    movdqa 1600(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm9[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm2[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 64(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm9[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 176(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm12[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm9[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 288(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm9[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm12[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 400(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm9[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm5[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 512(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm11[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm9[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 624(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm9[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm5[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 736(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm7[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm9[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm7[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm4[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 848(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm9[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 960(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm0[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 1072(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm14[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 1184(%rdi), %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm13[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,1,1]
+; SSE-NEXT:    movdqa %xmm2, %xmm13
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm15[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 1296(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 1408(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm6[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 1520(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm1[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 1632(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm8[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm14[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa 1744(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 96(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 208(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 320(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 432(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 544(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $250, (%rsp), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 656(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 768(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 880(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 992(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1104(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1216(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa %xmm13, %xmm7
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1328(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1440(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1552(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1664(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 1776(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm14[2],xmm0[3],xmm14[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
+; SSE-NEXT:    movdqa (%rsp), %xmm10 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm11, %xmm12
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm9, %xmm10
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm8, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm3, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm1[0],xmm14[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm1[0],xmm13[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm1[0],xmm12[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, (%rsp), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm1[0],xmm11[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm1[0],xmm10[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm1[0],xmm6[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r9)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movapd %xmm2, 240(%rax)
+; SSE-NEXT:    movapd %xmm3, 224(%rax)
+; SSE-NEXT:    movapd %xmm4, 208(%rax)
+; SSE-NEXT:    movapd %xmm5, 192(%rax)
+; SSE-NEXT:    movapd %xmm6, 176(%rax)
+; SSE-NEXT:    movapd %xmm7, 160(%rax)
+; SSE-NEXT:    movapd %xmm8, 144(%rax)
+; SSE-NEXT:    movapd %xmm9, 128(%rax)
+; SSE-NEXT:    movapd %xmm10, 112(%rax)
+; SSE-NEXT:    movapd %xmm11, 96(%rax)
+; SSE-NEXT:    movapd %xmm12, 80(%rax)
+; SSE-NEXT:    movapd %xmm13, 64(%rax)
+; SSE-NEXT:    movapd %xmm14, 48(%rax)
+; SSE-NEXT:    movapd %xmm15, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    addq $2456, %rsp # imm = 0x998
+; SSE-NEXT:    retq
+;
+; AVX1-ONLY-LABEL: load_i32_stride7_vf64:
+; AVX1-ONLY:       # %bb.0:
+; AVX1-ONLY-NEXT:    subq $3224, %rsp # imm = 0xC98
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovaps 304(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6],ymm1[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm9[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm7[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 752(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm0[0],ymm5[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm5
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovaps 1200(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm1[0],ymm4[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1280(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps 1312(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1600(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1568(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1648(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1728(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps 1760(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm13[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, %ymm10
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm3[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, %xmm15
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 528(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm12[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm11[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 976(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1024(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm8[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1424(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1504(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1472(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm6[1,1],ymm14[2,2],ymm6[5,5],ymm14[6,6]
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm9[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm7[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,1],ymm0[2,2],ymm1[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[0,0],ymm1[3,3],ymm5[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = zero,xmm1[1,2],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm4[1,1],ymm0[2,2],ymm4[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1280(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,0],ymm1[3,3],ymm4[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm4[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1632(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,1],ymm0[2,2],ymm1[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 1600(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1728(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = zero,xmm1[1,2],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm10[1,1],ymm0[2,2],ymm10[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm13[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm3[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm10[1,1],ymm0[2,2],ymm10[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm15[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[0,0],ymm2[3,3],ymm3[4,4],ymm2[7,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm2[1,2],xmm12[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,1],ymm0[2,2],ymm1[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0],xmm11[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,0],mem[3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1024(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm2[0,0],ymm6[3,3],ymm2[4,4],ymm6[7,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm6 = zero,xmm6[1,2],xmm8[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm12[1,1],ymm8[2,2],ymm12[5,5],ymm8[6,6]
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm6[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm11 = xmm0[0],xmm5[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[1,0],mem[3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm6[3,4],ymm11[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1472(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 1504(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm3[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm2[0,0],ymm15[3,3],ymm2[4,4],ymm15[7,7]
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, %ymm13
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm15 = zero,xmm15[1,2],xmm6[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm11[0],mem[1],xmm11[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm14[0,3],ymm15[7,5],ymm14[4,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm0[2,1],ymm15[2,0],ymm0[6,5],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1,2],xmm7[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0],xmm7[1],xmm11[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm15[3,1],mem[0,3],ymm15[7,5],mem[4,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm0[2,1],ymm15[2,0],ymm0[6,5],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm11[0],mem[1],xmm11[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm15[3,1],mem[0,3],ymm15[7,5],mem[4,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm0[2,1],ymm15[2,0],ymm0[6,5],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0],xmm2[1],xmm11[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm15[3,1],mem[0,3],ymm15[7,5],mem[4,7]
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[2,1],ymm15[2,0],ymm4[6,5],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm9[0],mem[0],ymm9[2],mem[2]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0],xmm1[1],xmm11[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm8[0,3],ymm15[7,5],ymm8[4,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm12[2,1],ymm15[2,0],ymm12[6,5],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm13[0],ymm3[0],ymm13[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm15[0,1,2],xmm6[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm11[0,1,2,3,4],ymm4[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = xmm4[0],mem[1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm11 = ymm11[3,1],mem[0,3],ymm11[7,5],mem[4,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm1[2,1],ymm11[2,0],ymm1[6,5],ymm11[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm11[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm11 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm11, %xmm11
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm11[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm11[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = xmm4[0],mem[1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[3,1],ymm13[0,3],ymm11[7,5],ymm13[4,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm10[2,1],ymm11[2,0],ymm10[6,5],ymm11[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm11[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm11 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm11, %xmm11
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm11[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm11[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm10[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0],xmm5[1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[3,1],ymm6[0,3],ymm11[7,5],ymm6[4,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm12[2,1],ymm11[2,0],ymm12[6,5],ymm11[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm11[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm11 = ymm14[0],ymm1[0],ymm14[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm11, %xmm11
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm11[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm11[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm3[1,0],ymm11[0,0],ymm3[5,4],ymm11[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm11[3,1],ymm4[0,2],ymm11[7,5],ymm4[4,6]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm3[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[3,2,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm11[0,1],ymm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm8[0,1],mem[1,3],ymm8[4,5],mem[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm3[0,2],ymm15[2,0],ymm3[4,6],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[1,0],ymm11[0,0],ymm4[5,4],ymm11[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm11[3,1],ymm4[0,2],ymm11[7,5],ymm4[4,6]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm7[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[3,2,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm15[0,1],ymm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = ymm3[0,1],mem[1,3],ymm3[4,5],mem[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm3[0,2],ymm15[2,0],ymm3[4,6],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[1,0],ymm3[0,0],ymm4[5,4],ymm3[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm3[3,1],ymm4[0,2],ymm3[7,5],ymm4[4,6]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm3[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[3,2,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm15[0,1],ymm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1312(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm3[0,1],ymm0[1,3],ymm3[4,5],ymm0[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm0[0,2],ymm15[2,0],ymm0[4,6],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm3[1,0],ymm0[0,0],ymm3[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm0[3,1],ymm4[0,2],ymm0[7,5],ymm4[4,6]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = xmm2[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[3,2,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1760(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm0[0,1],mem[1,3],ymm0[4,5],mem[5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm9[0,2],ymm8[2,0],ymm9[4,6],ymm8[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm2[0,0],ymm0[5,4],ymm2[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[3,1],ymm0[0,2],ymm2[7,5],ymm0[4,6]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = mem[0,1,2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1536(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[0,1],ymm15[1,3],ymm2[4,5],ymm15[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[0,2],ymm3[2,0],ymm2[4,6],ymm3[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm6[1,0],ymm12[0,0],ymm6[5,4],ymm12[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm12[3,1],ymm0[0,2],ymm12[7,5],ymm0[4,6]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm5[0,1,2],xmm10[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %ymm11
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm11[0,1],ymm1[1,3],ymm11[4,5],ymm1[5,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, %ymm10
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm14[0,2],ymm3[2,0],ymm14[4,6],ymm3[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[1,0],ymm1[0,0],ymm13[5,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[3,1],ymm0[0,2],ymm1[7,5],ymm0[4,6]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm8[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm7[0,1],ymm2[1,3],ymm7[4,5],ymm2[5,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm6[0,2],ymm3[2,0],ymm6[4,6],ymm3[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm1[0,0],ymm0[5,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[3,1],ymm0[0,2],ymm1[7,5],ymm0[4,6]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm5[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm9[0,1],ymm12[1,3],ymm9[4,5],ymm12[5,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm0[0,2],ymm4[2,0],ymm0[4,6],ymm4[6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm0[3,0],ymm3[0,0],ymm0[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm9[1,0],ymm12[2,0],ymm9[5,4],ymm12[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[2,0],ymm3[6,4],ymm4[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm12[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = mem[0],xmm5[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm0[3,0],ymm3[0,0],ymm0[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = ymm9[1,0],mem[2,0],ymm9[5,4],mem[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[2,0],ymm3[6,4],ymm4[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = mem[0],xmm5[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[3,0],ymm3[0,0],ymm6[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm7[1,0],ymm2[2,0],ymm7[5,4],ymm2[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm3[2,0],ymm1[6,4],ymm3[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm5[0,1,2],xmm3[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm7[1,0],mem[2,0],ymm7[5,4],mem[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm3[2,0],ymm1[6,4],ymm3[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm0[0,1,2],xmm3[3]
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm14[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm14[3,0],ymm1[0,0],ymm14[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[1,0],ymm10[2,0],ymm11[5,4],ymm10[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,0],ymm1[6,4],ymm2[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = mem[0],xmm3[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm3[1,0],mem[2,0],ymm3[5,4],mem[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,0],ymm1[6,4],ymm2[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm0[0,1,2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = mem[0],xmm13[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm13[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[1,0],ymm15[2,0],ymm0[5,4],ymm15[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,0],ymm1[6,4],ymm2[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm2[0,1,2],xmm13[3]
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm6 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = mem[0],xmm6[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm0[1,0],mem[2,0],ymm0[5,4],mem[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm1[2,0],ymm6[2,0],ymm1[6,4],ymm6[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 1632(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm1[0,1,2],xmm13[3]
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = mem[0],xmm10[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm15[2,1],mem[3,3],ymm15[6,5],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm14[0],xmm8[1],xmm14[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm10[1,0],ymm6[2,0],ymm10[5,4],ymm6[6,4]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = mem[0,1,2],xmm12[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm12[0,0],ymm11[1,0],ymm12[4,4],ymm11[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm13[0,1],xmm10[3,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm9[2,1],mem[3,3],ymm9[6,5],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = mem[0],xmm13[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm10[1,0],ymm6[2,0],ymm10[5,4],ymm6[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm9 = mem[0,1,2],xmm9[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm10 = ymm10[0,0],mem[1,0],ymm10[4,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm10[0,1],xmm9[3,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm10[2,1],mem[3,3],ymm10[6,5],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm9 = mem[0],xmm9[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm9[1,0],ymm6[2,0],ymm9[5,4],ymm6[6,4]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = mem[0,1,2],xmm5[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm9 = ymm9[0,0],mem[1,0],ymm9[4,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm9, %xmm9
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[0,1],xmm5[3,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm5 = ymm7[2,1],mem[3,3],ymm7[6,5],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm6 = mem[0],xmm7[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[1,0],ymm5[2,0],ymm6[5,4],ymm5[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm6 = mem[0,1,2],xmm6[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm9 = ymm9[0,0],mem[1,0],ymm9[4,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm9, %xmm9
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm9[0,1],xmm6[3,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm5 = ymm5[2,1],mem[3,3],ymm5[6,5],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm6 = mem[0],xmm6[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[1,0],ymm5[2,0],ymm6[5,4],ymm5[6,4]
+; AVX1-ONLY-NEXT:    vblendps $7, (%rsp), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm6[0,0],mem[1,0],ymm6[4,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm6[0,1],xmm4[3,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = ymm3[2,1],mem[3,3],ymm3[6,5],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = mem[0],xmm9[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm5[1,0],ymm4[2,0],ymm5[5,4],ymm4[6,4]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = mem[0,1,2],xmm3[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm5 = ymm5[0,0],mem[1,0],ymm5[4,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm5[0,1],xmm3[3,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[2,1],mem[3,3],ymm3[6,5],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[0],xmm4[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[1,0],ymm3[2,0],ymm4[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = mem[0,1,2],xmm2[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = ymm4[0,0],mem[1,0],ymm4[4,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm4[0,1],xmm2[3,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm0[2,1],mem[3,3],ymm0[6,5],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = mem[0],xmm3[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[1,0],ymm2[2,0],ymm3[5,4],ymm2[6,4]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[0,0],mem[1,0],ymm3[4,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[0,1],xmm1[3,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm15[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm15[3,0],ymm1[0,0],ymm15[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm14[1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[1,0],ymm11[2,0],ymm12[5,4],ymm11[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[3,0],ymm1[0,0],ymm2[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm2[0],mem[1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,0],ymm2[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm2[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[1,0],mem[2,0],ymm3[5,4],mem[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm10[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[3,0],ymm1[0,0],ymm10[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[0],mem[1],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,0],ymm3[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = ymm4[1,0],mem[2,0],ymm4[5,4],mem[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[3,0],ymm3[0,0],ymm4[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = xmm4[0],mem[1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,0],ymm4[4,5],ymm3[6,4]
+; AVX1-ONLY-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = xmm4[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm5 = ymm5[1,0],mem[2,0],ymm5[5,4],mem[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm5[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm5[3,0],ymm3[0,0],ymm5[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = xmm5[0],mem[1],xmm5[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,0],ymm5[4,5],ymm3[6,4]
+; AVX1-ONLY-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps $8, (%rsp), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = xmm5[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm6[1,0],mem[2,0],ymm6[5,4],mem[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[3,0],ymm5[0,0],ymm6[7,4],ymm5[4,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm6 = xmm6[0],mem[1],xmm6[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,0],ymm6[4,5],ymm5[6,4]
+; AVX1-ONLY-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm6 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm6 = xmm6[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = ymm7[1,0],mem[2,0],ymm7[5,4],mem[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm7[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[3,0],ymm6[0,0],ymm7[7,4],ymm6[4,4]
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = xmm7[0],mem[1],xmm7[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,0],ymm7[4,5],ymm6[6,4]
+; AVX1-ONLY-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = xmm7[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm8[1,0],mem[2,0],ymm8[5,4],mem[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[3,0],ymm7[0,0],ymm8[7,4],ymm7[4,4]
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = xmm8[0],mem[1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,0],ymm8[4,5],ymm7[6,4]
+; AVX1-ONLY-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = xmm8[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm10 = ymm9[1,0],mem[2,0],ymm9[5,4],mem[6,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm10[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%r9)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rax)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm6, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%rax)
+; AVX1-ONLY-NEXT:    addq $3224, %rsp # imm = 0xC98
+; AVX1-ONLY-NEXT:    vzeroupper
+; AVX1-ONLY-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: load_i32_stride7_vf64:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    subq $2680, %rsp # imm = 0xA78
+; AVX2-SLOW-NEXT:    vmovdqa 1216(%rdi), %ymm12
+; AVX2-SLOW-NEXT:    vmovdqu %ymm12, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 1152(%rdi), %ymm4
+; AVX2-SLOW-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 1120(%rdi), %ymm5
+; AVX2-SLOW-NEXT:    vmovdqa 768(%rdi), %ymm13
+; AVX2-SLOW-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 704(%rdi), %ymm6
+; AVX2-SLOW-NEXT:    vmovdqa 672(%rdi), %ymm7
+; AVX2-SLOW-NEXT:    vmovdqa 320(%rdi), %ymm8
+; AVX2-SLOW-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 256(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovdqa 224(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm0 = <0,7,6,u>
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
+; AVX2-SLOW-NEXT:    vmovdqa %ymm3, %ymm11
+; AVX2-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa %ymm2, %ymm10
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT:    vpbroadcastq 304(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqa 352(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovdqa 384(%rdi), %xmm3
+; AVX2-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT:    vpbroadcastd 420(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm6[6],ymm7[7]
+; AVX2-SLOW-NEXT:    vmovdqa %ymm7, %ymm9
+; AVX2-SLOW-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa %ymm6, %ymm8
+; AVX2-SLOW-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT:    vpbroadcastq 752(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqa 800(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovdqa 832(%rdi), %xmm3
+; AVX2-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT:    vpbroadcastd 868(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT:    vpbroadcastq 1200(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqa 1248(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovdqa 1280(%rdi), %xmm3
+; AVX2-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT:    vpbroadcastd 1316(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 1600(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 1568(%rdi), %ymm12
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5],ymm1[6],ymm12[7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT:    vmovdqa 1664(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpbroadcastq 1648(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqa 1696(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovdqa 1728(%rdi), %xmm3
+; AVX2-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT:    vpbroadcastd 1764(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 96(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpbroadcastq 80(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 32(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
+; AVX2-SLOW-NEXT:    vpermd %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqa 128(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovdqa 160(%rdi), %xmm3
+; AVX2-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT:    vpbroadcastd 196(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 480(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovdqa 448(%rdi), %ymm13
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm2[6],ymm13[7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa %ymm2, %ymm15
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT:    vmovdqa 544(%rdi), %ymm6
+; AVX2-SLOW-NEXT:    vpbroadcastq 528(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqa 576(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovdqa 608(%rdi), %xmm3
+; AVX2-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT:    vpbroadcastd 644(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 928(%rdi), %ymm14
+; AVX2-SLOW-NEXT:    vmovdqa 896(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm14[6],ymm1[7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT:    vmovdqa 992(%rdi), %ymm7
+; AVX2-SLOW-NEXT:    vpbroadcastq 976(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqa 1024(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovdqa 1056(%rdi), %xmm3
+; AVX2-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT:    vpbroadcastd 1092(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 1376(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 1344(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermd %ymm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vmovdqa 1440(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpbroadcastq 1424(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqa 1472(%rdi), %xmm1
+; AVX2-SLOW-NEXT:    vmovdqa 1504(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT:    vpbroadcastd 1540(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm3
+; AVX2-SLOW-NEXT:    vmovdqa 384(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 352(%rdi), %ymm0
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10,11],ymm1[28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vmovdqa 288(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3],ymm4[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm10[0],ymm11[1],ymm10[2,3,4],ymm11[5],ymm10[6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vmovdqa 832(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 800(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vmovdqa 736(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3],ymm11[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm8[0],ymm9[1],ymm8[2,3,4],ymm9[5],ymm8[6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vmovdqa 1280(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 1248(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vmovdqa 1184(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqu (%rsp), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3],ymm9[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-SLOW-NEXT:    vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0],ymm5[1],mem[2,3,4],ymm5[5],mem[6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vmovdqa 1728(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 1696(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vmovdqa 1632(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3],ymm5[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-SLOW-NEXT:    vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0],ymm12[1],mem[2,3,4],ymm12[5],mem[6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vmovdqa 608(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 576(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vmovdqa 512(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3],ymm6[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm15[0],ymm13[1],ymm15[2,3,4],ymm13[5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vmovdqa 1056(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 1024(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vmovdqa 960(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3],ymm7[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-SLOW-NEXT:    vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vmovdqa 1504(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovdqa 1472(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT:    vmovdqa %ymm2, %ymm12
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vmovdqa 1408(%rdi), %ymm7
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm7[2,3],ymm15[4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vmovdqa 160(%rdi), %ymm13
+; AVX2-SLOW-NEXT:    vmovdqa 128(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm13[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm13[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT:    vmovdqa %ymm3, %ymm14
+; AVX2-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vmovdqa 64(%rdi), %ymm6
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm6[2,3],ymm8[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0],ymm2[1],mem[2,3,4],ymm2[5],mem[6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 304(%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm4[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vpbroadcastd 232(%rdi), %xmm1
+; AVX2-SLOW-NEXT:    vmovdqa 256(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} ymm1 = ymm10[0],ymm4[0],ymm10[2],ymm4[2]
+; AVX2-SLOW-NEXT:    vpbroadcastd 428(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 752(%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm11[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vpbroadcastd 680(%rdi), %xmm1
+; AVX2-SLOW-NEXT:    vmovdqa 704(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = ymm3[0],mem[0],ymm3[2],mem[2]
+; AVX2-SLOW-NEXT:    vpbroadcastd 876(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 1200(%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm9[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vpbroadcastd 1128(%rdi), %xmm1
+; AVX2-SLOW-NEXT:    vmovdqa 1152(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-SLOW-NEXT:    vpbroadcastd 1324(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 1648(%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm5[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vpbroadcastd 1576(%rdi), %xmm1
+; AVX2-SLOW-NEXT:    vmovdqa 1600(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-SLOW-NEXT:    vpbroadcastd 1772(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 1424(%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm15[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],ymm15[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vpbroadcastd 1352(%rdi), %xmm1
+; AVX2-SLOW-NEXT:    vmovdqa 1376(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm12[0],ymm1[2],ymm12[2]
+; AVX2-SLOW-NEXT:    vpbroadcastd 1548(%rdi), %ymm7
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 80(%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm8[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vpbroadcastd 8(%rdi), %xmm1
+; AVX2-SLOW-NEXT:    vmovdqa 32(%rdi), %xmm5
+; AVX2-SLOW-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} ymm1 = ymm14[0],ymm13[0],ymm14[2],ymm13[2]
+; AVX2-SLOW-NEXT:    vpbroadcastd 204(%rdi), %ymm15
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 528(%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vpbroadcastd 456(%rdi), %xmm15
+; AVX2-SLOW-NEXT:    vmovdqa 480(%rdi), %xmm1
+; AVX2-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm15 = xmm15[0],xmm1[1],xmm15[2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} ymm15 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
+; AVX2-SLOW-NEXT:    vpbroadcastd 652(%rdi), %ymm14
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqa 976(%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpalignr {{.*#+}} ymm14 = ymm9[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm14 = ymm14[0,1,2],ymm0[3],ymm14[4,5,6],ymm0[7]
+; AVX2-SLOW-NEXT:    vpbroadcastd 904(%rdi), %xmm15
+; AVX2-SLOW-NEXT:    vmovdqa 928(%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} ymm15 = ymm7[0],ymm1[0],ymm7[2],ymm1[2]
+; AVX2-SLOW-NEXT:    vpbroadcastd 1100(%rdi), %ymm13
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm13 = ymm5[0],mem[1],ymm5[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm5[0,1,2],mem[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[3,2,2,3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm10[0,2],ymm4[1,3],ymm10[4,6],ymm4[5,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 432(%rdi), %ymm14
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm6 = ymm4[0],mem[1],ymm4[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm5 = xmm4[0,1,2],mem[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm6 = ymm3[0,2],mem[1,3],ymm3[4,6],mem[5,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 880(%rdi), %ymm13
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups (%rsp), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm5 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],mem[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[3,2,2,3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm13[0,2],ymm14[1,3],ymm13[4,6],ymm14[5,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1328(%rdi), %ymm6
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm4 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm4 = ymm4[0,2],mem[1,3],ymm4[4,6],mem[5,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1776(%rdi), %ymm5
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],mem[3]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = ymm15[0,2],mem[1,3],ymm15[4,6],mem[5,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1552(%rdi), %ymm4
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm9[0],ymm8[1],ymm9[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[0,2],ymm1[1,3],ymm7[4,6],ymm1[5,7]
+; AVX2-SLOW-NEXT:    vmovaps %ymm1, %ymm10
+; AVX2-SLOW-NEXT:    vmovaps %ymm7, %ymm8
+; AVX2-SLOW-NEXT:    vbroadcastss 1104(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovdqa %ymm11, %ymm7
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[0,2],ymm11[1,3],ymm12[4,6],ymm11[5,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 656(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm4[1,3],ymm3[4,6],ymm4[5,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 208(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vbroadcastss 100(%rdi), %xmm1
+; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
+; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm5 = <4,3,u,u>
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm2, %ymm5, %ymm2
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vbroadcastsd {{.*#+}} ymm9 = [30064771072,30064771072,30064771072,30064771072]
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm9, %ymm2
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 212(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm5, %ymm2
+; AVX2-SLOW-NEXT:    vbroadcastss 324(%rdi), %xmm3
+; AVX2-SLOW-NEXT:    vmovaps 288(%rdi), %xmm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm1[0,1,2],xmm3[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 436(%rdi), %ymm4
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm2, %ymm5, %ymm3
+; AVX2-SLOW-NEXT:    vbroadcastss 548(%rdi), %xmm4
+; AVX2-SLOW-NEXT:    vmovaps 512(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-SLOW-NEXT:    vpermps %ymm12, %ymm9, %ymm4
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 660(%rdi), %ymm6
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm5, %ymm4
+; AVX2-SLOW-NEXT:    vbroadcastss 772(%rdi), %xmm6
+; AVX2-SLOW-NEXT:    vmovaps 736(%rdi), %xmm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,3]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 884(%rdi), %ymm7
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm4, %ymm5, %ymm6
+; AVX2-SLOW-NEXT:    vbroadcastss 996(%rdi), %xmm7
+; AVX2-SLOW-NEXT:    vmovaps 960(%rdi), %xmm4
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm7 = xmm4[0,1,2],xmm7[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
+; AVX2-SLOW-NEXT:    vpermps %ymm8, %ymm9, %ymm7
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1108(%rdi), %ymm8
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm5, %ymm6
+; AVX2-SLOW-NEXT:    vbroadcastss 1220(%rdi), %xmm8
+; AVX2-SLOW-NEXT:    vmovaps 1184(%rdi), %xmm7
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm8 = xmm7[0,1,2],xmm8[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3]
+; AVX2-SLOW-NEXT:    vpermps %ymm13, %ymm9, %ymm8
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1332(%rdi), %ymm10
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm10[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm5, %ymm6
+; AVX2-SLOW-NEXT:    vbroadcastss 1444(%rdi), %xmm8
+; AVX2-SLOW-NEXT:    vmovaps 1408(%rdi), %xmm13
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm8 = xmm13[0,1,2],xmm8[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3]
+; AVX2-SLOW-NEXT:    vpermps %ymm15, %ymm9, %ymm8
+; AVX2-SLOW-NEXT:    vmovaps %ymm15, %ymm14
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1556(%rdi), %ymm10
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm10[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm5, %ymm5
+; AVX2-SLOW-NEXT:    vbroadcastss 1668(%rdi), %xmm6
+; AVX2-SLOW-NEXT:    vmovaps 1632(%rdi), %xmm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
+; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1780(%rdi), %ymm8
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-SLOW-NEXT:    vbroadcastss 216(%rdi), %ymm6
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %xmm6
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm6[0,1,2],xmm0[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm8 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm8[0,1],xmm0[2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 320(%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm5 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm5, %xmm5
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm5 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-SLOW-NEXT:    vbroadcastss 440(%rdi), %ymm8
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 544(%rdi), %xmm5
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0,1,2],xmm2[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-SLOW-NEXT:    vbroadcastss 664(%rdi), %ymm8
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 768(%rdi), %xmm8
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm3[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-SLOW-NEXT:    vbroadcastss 888(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 992(%rdi), %xmm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm4[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-SLOW-NEXT:    vbroadcastss 1112(%rdi), %ymm4
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps 1216(%rdi), %xmm2
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm7[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm4 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm4, %xmm4
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-SLOW-NEXT:    vbroadcastss 1336(%rdi), %ymm7
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 1440(%rdi), %xmm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm13[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm7 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm7, %xmm7
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm12[0],ymm14[1],ymm12[2,3,4],ymm14[5],ymm12[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
+; AVX2-SLOW-NEXT:    vbroadcastss 1560(%rdi), %ymm13
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm4[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps 1664(%rdi), %xmm4
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm13 = xmm4[0,1,2],xmm15[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm13 = xmm15[0,1],xmm13[2,3]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,0,3]
+; AVX2-SLOW-NEXT:    vbroadcastss 1784(%rdi), %ymm12
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1,2,3,4,5,6],ymm12[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 136(%rdi), %xmm12
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT:    vpermps 192(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 80(%rdi), %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm15[0,1],xmm6[2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 360(%rdi), %xmm12
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT:    vpermps 416(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 304(%rdi), %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm15[0,1,2],xmm0[3]
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm15[0,1],xmm0[2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 584(%rdi), %xmm12
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT:    vpermps 640(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 528(%rdi), %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm5 = xmm15[0,1,2],xmm5[3]
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm5 = xmm15[0,1],xmm5[2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 808(%rdi), %xmm12
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT:    vpermps 864(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 752(%rdi), %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm14 = xmm15[0,1,2],xmm8[3]
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1032(%rdi), %xmm12
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT:    vpermps 1088(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 976(%rdi), %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm15[0,1,2],xmm1[3]
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1256(%rdi), %xmm12
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT:    vpermps 1312(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1200(%rdi), %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm15[0,1,2],xmm2[3]
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm15[0,1],xmm2[2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1480(%rdi), %xmm12
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT:    vpermps 1536(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1424(%rdi), %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm15[0,1,2],xmm3[3]
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm15[0,1],xmm3[2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1704(%rdi), %xmm12
+; AVX2-SLOW-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT:    vpermps 1760(%rdi), %ymm9, %ymm9
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm12[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 1648(%rdi), %ymm12
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm12[0,1,2],xmm4[3]
+; AVX2-SLOW-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm12 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm12, %xmm12
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm12[0,1],xmm4[2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 192(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 128(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 64(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, (%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 224(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 160(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 96(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 32(%rsi)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 192(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 128(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 64(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, (%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 224(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 160(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 96(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 32(%rdx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 128(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 64(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, (%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 192(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 224(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 160(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 96(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 32(%rcx)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, (%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 64(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 128(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 192(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 224(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 160(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 96(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 32(%r8)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm8, 224(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm8, 192(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm8, 160(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm8, 128(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm8, 96(%r9)
+; AVX2-SLOW-NEXT:    vmovups (%rsp), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 64(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 32(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, (%r9)
+; AVX2-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-SLOW-NEXT:    vmovaps %ymm13, 224(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm7, 192(%rax)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm7, 160(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm11, 128(%rax)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm7, 96(%rax)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm7, 64(%rax)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm7, 32(%rax)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm7, (%rax)
+; AVX2-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-SLOW-NEXT:    vmovaps %ymm4, 224(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm3, 192(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm2, 160(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm1, 128(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm14, 96(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm5, 64(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm6, (%rax)
+; AVX2-SLOW-NEXT:    addq $2680, %rsp # imm = 0xA78
+; AVX2-SLOW-NEXT:    vzeroupper
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: load_i32_stride7_vf64:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    subq $2680, %rsp # imm = 0xA78
+; AVX2-FAST-NEXT:    vmovdqa 1216(%rdi), %ymm8
+; AVX2-FAST-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 1152(%rdi), %ymm11
+; AVX2-FAST-NEXT:    vmovdqa 1120(%rdi), %ymm5
+; AVX2-FAST-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 768(%rdi), %ymm7
+; AVX2-FAST-NEXT:    vmovdqa 704(%rdi), %ymm6
+; AVX2-FAST-NEXT:    vmovdqa 672(%rdi), %ymm14
+; AVX2-FAST-NEXT:    vmovdqa 320(%rdi), %ymm10
+; AVX2-FAST-NEXT:    vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 256(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovdqa 224(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} xmm0 = <0,7,6,u>
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
+; AVX2-FAST-NEXT:    vmovdqa %ymm3, %ymm13
+; AVX2-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa %ymm2, %ymm12
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-NEXT:    vpbroadcastq 304(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqa 352(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovdqa 384(%rdi), %xmm3
+; AVX2-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpbroadcastd 420(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm6[6],ymm14[7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa %ymm6, %ymm10
+; AVX2-FAST-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-NEXT:    vpbroadcastq 752(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqa 800(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovdqa 832(%rdi), %xmm3
+; AVX2-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpbroadcastd 868(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm11[6],ymm5[7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-NEXT:    vpbroadcastq 1200(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqa 1248(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovdqa 1280(%rdi), %xmm3
+; AVX2-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpbroadcastd 1316(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 1600(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 1568(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-NEXT:    vmovdqa 1664(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vpbroadcastq 1648(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovdqa %ymm3, %ymm6
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqa 1696(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovdqa 1728(%rdi), %xmm3
+; AVX2-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpbroadcastd 1764(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 96(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpbroadcastq 80(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovdqa (%rdi), %ymm3
+; AVX2-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 32(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
+; AVX2-FAST-NEXT:    vpermd %ymm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqa 128(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovdqa 160(%rdi), %xmm3
+; AVX2-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpbroadcastd 196(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 480(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 448(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-NEXT:    vmovdqa 544(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpbroadcastq 528(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqa 576(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovdqa 608(%rdi), %xmm3
+; AVX2-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpbroadcastd 644(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 928(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovdqa 896(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
+; AVX2-FAST-NEXT:    vmovdqa %ymm2, %ymm15
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-NEXT:    vmovdqa 992(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpbroadcastq 976(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqa 1024(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovdqa 1056(%rdi), %xmm3
+; AVX2-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vpbroadcastd 1092(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 1376(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 1344(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vmovdqa 1440(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpbroadcastq 1424(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqa 1472(%rdi), %xmm1
+; AVX2-FAST-NEXT:    vmovdqa 1504(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-FAST-NEXT:    vpbroadcastd 1540(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vmovdqa 384(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 352(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vmovdqa 288(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3],ymm8[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [25769803781,25769803781,25769803781,25769803781]
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm5 = [1,0,7,7,5,4,7,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm5, %ymm4
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vmovdqa 832(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 800(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm3 = ymm0[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm0[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vmovdqa 736(%rdi), %ymm12
+; AVX2-FAST-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm12[2,3],ymm7[4,5],ymm12[6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm10[0],ymm14[1],ymm10[2,3,4],ymm14[5],ymm10[6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm5, %ymm4
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vmovdqa 1280(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 1248(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm3 = ymm3[12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10,11],ymm3[28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vmovdqa 1184(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm0[2,3],ymm9[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm5, %ymm4
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vmovdqa 1728(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 1696(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm3 = ymm0[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm0[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vmovdqa 1632(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1],ymm0[2,3],ymm6[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovdqa %ymm6, %ymm14
+; AVX2-FAST-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm4 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm4, %ymm5, %ymm4
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vmovdqa 160(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 128(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm3 = ymm0[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm0[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vmovdqa 64(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm11[0,1],ymm0[2,3],ymm11[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm4 = mem[0],ymm0[1],mem[2,3,4],ymm0[5],mem[6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm4, %ymm5, %ymm4
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vmovdqa 608(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 576(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm3 = ymm0[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm0[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vmovdqa 512(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm13[0,1],ymm0[2,3],ymm13[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm5, %ymm4
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vmovdqa 1056(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 1024(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm3 = ymm0[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm0[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vmovdqa 960(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermd %ymm3, %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm5, %ymm4
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT:    vmovdqa 1504(%rdi), %ymm0
+; AVX2-FAST-NEXT:    vmovdqa 1472(%rdi), %ymm10
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm3 = ymm0[12,13,14,15],ymm10[0,1,2,3,4,5,6,7,8,9,10,11],ymm0[28,29,30,31],ymm10[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-NEXT:    vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa %ymm0, %ymm15
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vmovdqa 1408(%rdi), %ymm6
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1],ymm6[2,3],ymm4[4,5],ymm6[6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm3, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm5, %ymm1
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 304(%rdi), %xmm0
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm1 = ymm8[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vpbroadcastd 232(%rdi), %xmm1
+; AVX2-FAST-NEXT:    vmovdqa 256(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
+; AVX2-FAST-NEXT:    vpbroadcastd 428(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 752(%rdi), %xmm0
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm1 = ymm7[8,9,10,11,12,13,14,15],ymm12[0,1,2,3,4,5,6,7],ymm7[24,25,26,27,28,29,30,31],ymm12[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vpbroadcastd 680(%rdi), %xmm1
+; AVX2-FAST-NEXT:    vmovdqa 704(%rdi), %xmm5
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = ymm12[0],mem[0],ymm12[2],mem[2]
+; AVX2-FAST-NEXT:    vpbroadcastd 876(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 1200(%rdi), %xmm0
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm1 = ymm9[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vpbroadcastd 1128(%rdi), %xmm1
+; AVX2-FAST-NEXT:    vmovdqa 1152(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-FAST-NEXT:    vpbroadcastd 1324(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 1648(%rdi), %xmm0
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm1 = ymm14[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm14[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vpbroadcastd 1576(%rdi), %xmm1
+; AVX2-FAST-NEXT:    vmovdqa 1600(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-FAST-NEXT:    vpbroadcastd 1772(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 1424(%rdi), %xmm0
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm1 = ymm4[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vpbroadcastd 1352(%rdi), %xmm1
+; AVX2-FAST-NEXT:    vmovdqa 1376(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpunpcklqdq {{.*#+}} ymm1 = ymm10[0],ymm15[0],ymm10[2],ymm15[2]
+; AVX2-FAST-NEXT:    vpbroadcastd 1548(%rdi), %ymm7
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 80(%rdi), %xmm0
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm1 = ymm11[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vpbroadcastd 8(%rdi), %xmm1
+; AVX2-FAST-NEXT:    vmovdqa 32(%rdi), %xmm4
+; AVX2-FAST-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpunpcklqdq (%rsp), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-FAST-NEXT:    vpbroadcastd 204(%rdi), %ymm15
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 528(%rdi), %xmm0
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm1 = ymm13[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm13[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vpbroadcastd 456(%rdi), %xmm15
+; AVX2-FAST-NEXT:    vmovdqa 480(%rdi), %xmm1
+; AVX2-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm15 = xmm15[0],xmm1[1],xmm15[2,3]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = ymm9[0],mem[0],ymm9[2],mem[2]
+; AVX2-FAST-NEXT:    vpbroadcastd 652(%rdi), %ymm14
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqa 976(%rdi), %xmm0
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpalignr {{.*#+}} ymm14 = ymm10[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm14 = ymm14[0,1,2],ymm0[3],ymm14[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vpbroadcastd 904(%rdi), %xmm15
+; AVX2-FAST-NEXT:    vmovdqa 928(%rdi), %xmm0
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpunpcklqdq {{.*#+}} ymm15 = ymm7[0],ymm1[0],ymm7[2],ymm1[2]
+; AVX2-FAST-NEXT:    vpbroadcastd 1100(%rdi), %ymm13
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm11 = ymm14[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm13 = ymm4[0],mem[1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm6 = xmm4[0,1,2],mem[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[3,2,2,3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm13 = ymm15[0,2],ymm3[1,3],ymm15[4,6],ymm3[5,7]
+; AVX2-FAST-NEXT:    vbroadcastss 432(%rdi), %ymm14
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm6 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1,2],mem[3]
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} ymm6 = ymm6[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm6 = ymm12[0,2],ymm14[1,3],ymm12[4,6],ymm14[5,7]
+; AVX2-FAST-NEXT:    vbroadcastss 880(%rdi), %ymm12
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm12[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm5 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],mem[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[3,2,2,3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm5 = ymm3[0,2],mem[1,3],ymm3[4,6],mem[5,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1328(%rdi), %ymm6
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm4 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm4 = ymm4[0,2],mem[1,3],ymm4[4,6],mem[5,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1776(%rdi), %ymm5
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],mem[3]
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3]
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = ymm12[0,2],mem[1,3],ymm12[4,6],mem[5,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1552(%rdi), %ymm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm10[0],ymm8[1],ymm10[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3]
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3]
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[0,2],ymm1[1,3],ymm7[4,6],ymm1[5,7]
+; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm11
+; AVX2-FAST-NEXT:    vmovaps %ymm7, %ymm8
+; AVX2-FAST-NEXT:    vbroadcastss 1104(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[0,2],ymm7[1,3],ymm9[4,6],ymm7[5,7]
+; AVX2-FAST-NEXT:    vbroadcastss 656(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovups (%rsp), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm4[1,3],ymm3[4,6],ymm4[5,7]
+; AVX2-FAST-NEXT:    vbroadcastss 208(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vbroadcastss 100(%rdi), %xmm1
+; AVX2-FAST-NEXT:    vmovaps 64(%rdi), %xmm0
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm5 = <4,3,u,u>
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm5, %ymm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-FAST-NEXT:    vbroadcastsd {{.*#+}} ymm10 = [30064771072,30064771072,30064771072,30064771072]
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm10, %ymm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 212(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm5, %ymm2
+; AVX2-FAST-NEXT:    vbroadcastss 324(%rdi), %xmm3
+; AVX2-FAST-NEXT:    vmovaps 288(%rdi), %xmm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm1[0,1,2],xmm3[3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
+; AVX2-FAST-NEXT:    vpermps %ymm15, %ymm10, %ymm3
+; AVX2-FAST-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 436(%rdi), %ymm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm5, %ymm3
+; AVX2-FAST-NEXT:    vbroadcastss 548(%rdi), %xmm4
+; AVX2-FAST-NEXT:    vmovaps 512(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm10, %ymm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 660(%rdi), %ymm6
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm5, %ymm4
+; AVX2-FAST-NEXT:    vbroadcastss 772(%rdi), %xmm6
+; AVX2-FAST-NEXT:    vmovaps 736(%rdi), %xmm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,3]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermps %ymm15, %ymm10, %ymm6
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 884(%rdi), %ymm7
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm5, %ymm6
+; AVX2-FAST-NEXT:    vbroadcastss 996(%rdi), %xmm7
+; AVX2-FAST-NEXT:    vmovaps 960(%rdi), %xmm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm7 = xmm4[0,1,2],xmm7[3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
+; AVX2-FAST-NEXT:    vpermps %ymm8, %ymm10, %ymm7
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1108(%rdi), %ymm8
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm5, %ymm6
+; AVX2-FAST-NEXT:    vbroadcastss 1220(%rdi), %xmm8
+; AVX2-FAST-NEXT:    vmovaps 1184(%rdi), %xmm7
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm8 = xmm7[0,1,2],xmm8[3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3]
+; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = ymm8[0,1,2,3,4,5],mem[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1332(%rdi), %ymm9
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm5, %ymm6
+; AVX2-FAST-NEXT:    vbroadcastss 1444(%rdi), %xmm9
+; AVX2-FAST-NEXT:    vmovaps 1408(%rdi), %xmm8
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm9 = xmm8[0,1,2],xmm9[3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm9[2,3]
+; AVX2-FAST-NEXT:    vpermps %ymm12, %ymm10, %ymm9
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1556(%rdi), %ymm11
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5,6],ymm11[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm5, %ymm5
+; AVX2-FAST-NEXT:    vbroadcastss 1668(%rdi), %xmm6
+; AVX2-FAST-NEXT:    vmovaps 1632(%rdi), %xmm11
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm6 = xmm11[0,1,2],xmm6[3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT:    vpermps %ymm13, %ymm10, %ymm6
+; AVX2-FAST-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1780(%rdi), %ymm9
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm9[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups (%rsp), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm14 = [1,0,3,3,1,0,7,7]
+; AVX2-FAST-NEXT:    vpermps %ymm5, %ymm14, %ymm5
+; AVX2-FAST-NEXT:    vbroadcastss 216(%rdi), %ymm6
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %xmm9
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm9[0,1,2],xmm0[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm6 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm6, %xmm6
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm6[0,1],xmm0[2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 320(%rdi), %xmm6
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm6[0,1,2],xmm1[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[0],ymm1[1],mem[2,3,4],ymm1[5],mem[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm14, %ymm1
+; AVX2-FAST-NEXT:    vbroadcastss 440(%rdi), %ymm5
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 544(%rdi), %xmm5
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm5[0,1,2],xmm2[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm14, %ymm1
+; AVX2-FAST-NEXT:    vbroadcastss 664(%rdi), %ymm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 768(%rdi), %xmm0
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm3[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-FAST-NEXT:    vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = mem[0],ymm15[1],mem[2,3,4],ymm15[5],mem[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm14, %ymm2
+; AVX2-FAST-NEXT:    vbroadcastss 888(%rdi), %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 992(%rdi), %xmm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm4[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,3,2]
+; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm3, %xmm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm14, %ymm3
+; AVX2-FAST-NEXT:    vbroadcastss 1112(%rdi), %ymm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 1216(%rdi), %xmm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm7[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,3,2]
+; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm4 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm4, %xmm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm14, %ymm4
+; AVX2-FAST-NEXT:    vbroadcastss 1336(%rdi), %ymm7
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovaps 1440(%rdi), %xmm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm8[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm4 = xmm8[0,1],xmm4[2,3]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm8, %ymm14, %ymm8
+; AVX2-FAST-NEXT:    vbroadcastss 1560(%rdi), %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovaps 1664(%rdi), %xmm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm11 = xmm4[0,1,2],xmm11[3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm11 = xmm15[0,1],xmm11[2,3]
+; AVX2-FAST-NEXT:    vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = mem[0],ymm13[1],mem[2,3,4],ymm13[5],mem[6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm15, %ymm14, %ymm14
+; AVX2-FAST-NEXT:    vbroadcastss 1784(%rdi), %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 136(%rdi), %xmm14
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm14 = xmm14[0],mem[1],xmm14[2,3]
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-FAST-NEXT:    vpermps 192(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 80(%rdi), %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm9 = xmm15[0,1,2],xmm9[3]
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm9 = xmm15[0,1],xmm9[2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 360(%rdi), %xmm14
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm14 = xmm14[0],mem[1],xmm14[2,3]
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-FAST-NEXT:    vpermps 416(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 304(%rdi), %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm13 = xmm15[0,1,2],xmm6[3]
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm13 = xmm15[0,1],xmm13[2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 584(%rdi), %xmm14
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm14 = xmm14[0],mem[1],xmm14[2,3]
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-FAST-NEXT:    vpermps 640(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 528(%rdi), %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm12 = xmm15[0,1,2],xmm5[3]
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm12 = xmm15[0,1],xmm12[2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm12[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 808(%rdi), %xmm12
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-NEXT:    vpermps 864(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 752(%rdi), %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm15[0,1,2],xmm0[3]
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm15[0,1],xmm0[2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1032(%rdi), %xmm12
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-NEXT:    vpermps 1088(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 976(%rdi), %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm15[0,1,2],xmm1[3]
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1256(%rdi), %xmm12
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-NEXT:    vpermps 1312(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1200(%rdi), %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm15[0,1,2],xmm2[3]
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm15[0,1],xmm2[2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1480(%rdi), %xmm12
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-NEXT:    vpermps 1536(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1424(%rdi), %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm15[0,1,2],xmm3[3]
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm15[0,1],xmm3[2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1704(%rdi), %xmm12
+; AVX2-FAST-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-NEXT:    vpermps 1760(%rdi), %ymm10, %ymm10
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss 1648(%rdi), %ymm12
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm4 = xmm12[0,1,2],xmm4[3]
+; AVX2-FAST-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm12 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
+; AVX2-FAST-NEXT:    vextractf128 $1, %ymm12, %xmm12
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm4 = xmm12[0,1],xmm4[2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 192(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 128(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 64(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, (%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 224(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 160(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 96(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 32(%rsi)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 192(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 128(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 64(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, (%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 224(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 160(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 96(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 32(%rdx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 128(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 64(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, (%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 192(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 224(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 160(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 96(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 32(%rcx)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, (%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 64(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 128(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 192(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 224(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 160(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 96(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 32(%r8)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm5, 224(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm5, 192(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm5, 160(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm5, 128(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm5, 96(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 64(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, 32(%r9)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm10, (%r9)
+; AVX2-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-FAST-NEXT:    vmovaps %ymm11, 224(%rax)
+; AVX2-FAST-NEXT:    vmovaps %ymm8, 192(%rax)
+; AVX2-FAST-NEXT:    vmovaps %ymm7, 160(%rax)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm5, 128(%rax)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm5, 96(%rax)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm5, 64(%rax)
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm5, 32(%rax)
+; AVX2-FAST-NEXT:    vmovups (%rsp), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovaps %ymm5, (%rax)
+; AVX2-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-FAST-NEXT:    vmovaps %ymm4, 224(%rax)
+; AVX2-FAST-NEXT:    vmovaps %ymm3, 192(%rax)
+; AVX2-FAST-NEXT:    vmovaps %ymm2, 160(%rax)
+; AVX2-FAST-NEXT:    vmovaps %ymm1, 128(%rax)
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 96(%rax)
+; AVX2-FAST-NEXT:    vmovaps %ymm14, 64(%rax)
+; AVX2-FAST-NEXT:    vmovaps %ymm13, 32(%rax)
+; AVX2-FAST-NEXT:    vmovaps %ymm9, (%rax)
+; AVX2-FAST-NEXT:    addq $2680, %rsp # imm = 0xA78
+; AVX2-FAST-NEXT:    vzeroupper
+; AVX2-FAST-NEXT:    retq
+;
+; AVX2-FAST-PERLANE-LABEL: load_i32_stride7_vf64:
+; AVX2-FAST-PERLANE:       # %bb.0:
+; AVX2-FAST-PERLANE-NEXT:    subq $2680, %rsp # imm = 0xA78
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1216(%rdi), %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm12, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1152(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1120(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 768(%rdi), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 704(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 672(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 320(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 256(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 224(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm0 = <0,7,6,u>
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm3, %ymm11
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm2, %ymm10
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq 304(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 352(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 384(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 420(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm6[6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm7, %ymm9
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm6, %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq 752(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 800(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 832(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 868(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq 1200(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1248(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1280(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 1316(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1600(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1568(%rdi), %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5],ymm1[6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1664(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq 1648(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1696(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1728(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 1764(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 96(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq 80(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 128(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 160(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 196(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 480(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 448(%rdi), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm2[6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm2, %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 544(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq 528(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 576(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 608(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 644(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 928(%rdi), %ymm14
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 896(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm14[6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 992(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq 976(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1024(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1056(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 1092(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1376(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1344(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm1, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1440(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq 1424(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1472(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1504(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 1540(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 384(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 352(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10,11],ymm1[28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 288(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3],ymm4[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm10[0],ymm11[1],ymm10[2,3,4],ymm11[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 832(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 800(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 736(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3],ymm11[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm8[0],ymm9[1],ymm8[2,3,4],ymm9[5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1280(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1248(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1184(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu (%rsp), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3],ymm9[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0],ymm5[1],mem[2,3,4],ymm5[5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1728(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1696(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1632(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3],ymm5[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0],ymm12[1],mem[2,3,4],ymm12[5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 608(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 576(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 512(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3],ymm6[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm15[0],ymm13[1],ymm15[2,3,4],ymm13[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1056(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1024(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 960(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3],ymm7[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1504(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1472(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm2, %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1408(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm7[2,3],ymm15[4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 160(%rdi), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 128(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm13[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm13[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm3, %ymm14
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 64(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm6[2,3],ymm8[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0],ymm2[1],mem[2,3,4],ymm2[5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 304(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm4[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 232(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 256(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpunpcklqdq {{.*#+}} ymm1 = ymm10[0],ymm4[0],ymm10[2],ymm4[2]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 428(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 752(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm11[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 680(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 704(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm3[0],mem[0],ymm3[2],mem[2]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 876(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1200(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm9[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 1128(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1152(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 1324(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1648(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm5[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 1576(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1600(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 1772(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1424(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm15[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],ymm15[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 1352(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 1376(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm12[0],ymm1[2],ymm12[2]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 1548(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 80(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm1 = ymm8[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 8(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpunpcklqdq {{.*#+}} ymm1 = ymm14[0],ymm13[0],ymm14[2],ymm13[2]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 204(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 528(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 456(%rdi), %xmm15
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 480(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm15 = xmm15[0],xmm1[1],xmm15[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpunpcklqdq {{.*#+}} ymm15 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 652(%rdi), %ymm14
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 976(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpalignr {{.*#+}} ymm14 = ymm9[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm14 = ymm14[0,1,2],ymm0[3],ymm14[4,5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 904(%rdi), %xmm15
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa 928(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpunpcklqdq {{.*#+}} ymm15 = ymm7[0],ymm1[0],ymm7[2],ymm1[2]
+; AVX2-FAST-PERLANE-NEXT:    vpbroadcastd 1100(%rdi), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm13 = ymm5[0],mem[1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm5[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm10[0,2],ymm4[1,3],ymm10[4,6],ymm4[5,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 432(%rdi), %ymm14
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm6 = ymm4[0],mem[1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm5 = xmm4[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm6 = ymm3[0,2],mem[1,3],ymm3[4,6],mem[5,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 880(%rdi), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups (%rsp), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm13[0,2],ymm14[1,3],ymm13[4,6],ymm14[5,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1328(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm4 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm4 = ymm4[0,2],mem[1,3],ymm4[4,6],mem[5,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1776(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = ymm15[0,2],mem[1,3],ymm15[4,6],mem[5,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1552(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm9[0],ymm8[1],ymm9[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[0,2],ymm1[1,3],ymm7[4,6],ymm1[5,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, %ymm10
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm7, %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1104(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm11, %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[0,2],ymm11[1,3],ymm12[4,6],ymm11[5,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 656(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm4[1,3],ymm3[4,6],ymm4[5,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 208(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 100(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm5 = <4,3,u,u>
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm2, %ymm5, %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{.*#+}} ymm9 = [30064771072,30064771072,30064771072,30064771072]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm9, %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 212(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm5, %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 324(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 288(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm1[0,1,2],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 436(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm2, %ymm5, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 548(%rdi), %xmm4
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 512(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm12, %ymm9, %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 660(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm5, %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 772(%rdi), %xmm6
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 736(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 884(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm4, %ymm5, %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 996(%rdi), %xmm7
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 960(%rdi), %xmm4
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm7 = xmm4[0,1,2],xmm7[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm8, %ymm9, %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1108(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm5, %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1220(%rdi), %xmm8
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1184(%rdi), %xmm7
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm8 = xmm7[0,1,2],xmm8[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm13, %ymm9, %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1332(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm5, %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1444(%rdi), %xmm8
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1408(%rdi), %xmm13
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm8 = xmm13[0,1,2],xmm8[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm15, %ymm9, %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm15, %ymm14
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1556(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm5, %ymm5
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1668(%rdi), %xmm6
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1632(%rdi), %xmm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1780(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 216(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %xmm6
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm6[0,1,2],xmm0[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm8 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm8[0,1],xmm0[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 320(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm5 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 440(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 544(%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0,1,2],xmm2[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 664(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 768(%rdi), %xmm8
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 888(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 992(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm3, %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1112(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1216(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm7[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm4 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1336(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1440(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm13[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm7 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm7, %xmm7
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm12[0],ymm14[1],ymm12[2,3,4],ymm14[5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1560(%rdi), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm4[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 1664(%rdi), %xmm4
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm13 = xmm4[0,1,2],xmm15[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm13 = xmm15[0,1],xmm13[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1784(%rdi), %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1,2,3,4,5,6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 136(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vpermps 192(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 80(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm15[0,1],xmm6[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 360(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vpermps 416(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 304(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm15[0,1,2],xmm0[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm15[0,1],xmm0[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 584(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vpermps 640(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 528(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm5 = xmm15[0,1,2],xmm5[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm5 = xmm15[0,1],xmm5[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 808(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vpermps 864(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 752(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm14 = xmm15[0,1,2],xmm8[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1032(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vpermps 1088(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 976(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm15[0,1,2],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1256(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vpermps 1312(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1200(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm15[0,1,2],xmm2[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm15[0,1],xmm2[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1480(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vpermps 1536(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1424(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm15[0,1,2],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm15[0,1],xmm3[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1704(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermps 1760(%rdi), %ymm9, %ymm9
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm12[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1648(%rdi), %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm12[0,1,2],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm12 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm12, %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm12[0,1],xmm4[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 192(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 128(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 64(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, (%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 224(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 160(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 96(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 32(%rsi)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 192(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 128(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 64(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, (%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 224(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 160(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 96(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 32(%rdx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 128(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 64(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, (%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 192(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 224(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 160(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 96(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, (%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 64(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 128(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 192(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 224(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 160(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 96(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 32(%r8)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, 224(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, 192(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, 160(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, 128(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, 96(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups (%rsp), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 64(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 32(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, (%r9)
+; AVX2-FAST-PERLANE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm13, 224(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm7, 192(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm7, 160(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm11, 128(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm7, 96(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm7, 64(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm7, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm7, (%rax)
+; AVX2-FAST-PERLANE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm4, 224(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm3, 192(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, 160(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, 128(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm14, 96(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm5, 64(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm6, (%rax)
+; AVX2-FAST-PERLANE-NEXT:    addq $2680, %rsp # imm = 0xA78
+; AVX2-FAST-PERLANE-NEXT:    vzeroupper
+; AVX2-FAST-PERLANE-NEXT:    retq
+;
+; AVX512F-LABEL: load_i32_stride7_vf64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    subq $3400, %rsp # imm = 0xD48
+; AVX512F-NEXT:    vmovdqa64 1728(%rdi), %zmm13
+; AVX512F-NEXT:    vmovdqa64 1664(%rdi), %zmm14
+; AVX512F-NEXT:    vmovdqa64 1600(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqa64 1280(%rdi), %zmm10
+; AVX512F-NEXT:    vmovdqa64 1216(%rdi), %zmm12
+; AVX512F-NEXT:    vmovdqa64 1152(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqa64 832(%rdi), %zmm17
+; AVX512F-NEXT:    vmovdqa64 768(%rdi), %zmm18
+; AVX512F-NEXT:    vmovdqa64 704(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 384(%rdi), %zmm15
+; AVX512F-NEXT:    vmovdqa64 320(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa64 256(%rdi), %zmm8
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [20,27,6,13,20,27,6,13,20,27,6,13,20,27,6,13]
+; AVX512F-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm2, %zmm3
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,10,11,12,13,18,25,0,0,10,11,12,13,18,25]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm12, %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm14, %zmm4, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,7,14,21,28,0,0,0,0,7,14,21,28,0,0]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm12, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,19,26,0,9,10,11,12,13,19,26]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm14, %zmm4, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,1,8,15,22,29,0,0,0,1,8,15,22,29,0,0]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm12, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,20,27,0,9,10,11,12,13,20,27]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm14, %zmm4, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,18,25,0,7,14,0,0,0,18,25,0,7,14,0,0]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm5, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,21,28,0,9,10,11,12,13,21,28]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm6, %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm8, %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm4, %zmm14, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,19,26,1,8,15,0,0,0,19,26,1,8,15,0,0]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm5, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,22,29,0,9,10,11,12,13,22,29]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm6, %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm8, %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm4, %zmm14, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1024(%rdi), %zmm11
+; AVX512F-NEXT:    vmovdqa64 1088(%rdi), %zmm16
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm22 = [25,4,11,18,25,4,11,18,25,4,11,18,25,4,11,18]
+; AVX512F-NEXT:    # zmm22 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm16, %zmm22, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %zmm19
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm22, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 128(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqa64 192(%rdi), %zmm20
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm22, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1472(%rdi), %zmm9
+; AVX512F-NEXT:    vmovdqa64 1536(%rdi), %zmm21
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm21, %zmm22, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm12, %zmm22, %zmm7
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,16,23,30,0,9,10,11,12,16,23,30]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm0, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm22, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm0, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm22, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm0, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm14, %zmm4, %zmm22
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [26,5,12,19,26,5,12,19,26,5,12,19,26,5,12,19]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm0, %zmm7
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,9,10,11,12,17,24,31,0,9,10,11,12,17,24,31]
+; AVX512F-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm6, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm0, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm6, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm14, %zmm0, %zmm4
+; AVX512F-NEXT:    vpermt2d %zmm13, %zmm6, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm12, %zmm0, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm10, %zmm6, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm16, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm21, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm24 = [8,15,0,0,0,19,26,1,8,15,0,0,0,19,26,1]
+; AVX512F-NEXT:    # zmm24 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm19, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm3, %zmm24, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm27 = [27,6,13,20,27,6,13,20,27,6,13,20,27,6,13,20]
+; AVX512F-NEXT:    # zmm27 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm27, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm28 = [28,0,0,0,0,7,14,21,28,0,0,0,0,7,14,21]
+; AVX512F-NEXT:    # zmm28 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm28, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm29 = [29,0,0,0,1,8,15,22,29,0,0,0,1,8,15,22]
+; AVX512F-NEXT:    # zmm29 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm29, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [14,0,0,0,18,25,0,7,14,0,0,0,18,25,0,7]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm3, %zmm0, %zmm19
+; AVX512F-NEXT:    vmovdqu64 %zmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm24, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm27, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm28, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm29, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm0, %zmm20
+; AVX512F-NEXT:    vmovdqu64 %zmm20, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm24, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm9, %zmm21, %zmm24
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm16, %zmm27, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm21, %zmm9, %zmm27
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm16, %zmm28, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm21, %zmm9, %zmm28
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm16, %zmm29, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm21, %zmm9, %zmm29
+; AVX512F-NEXT:    vpermt2d %zmm9, %zmm0, %zmm21
+; AVX512F-NEXT:    vmovdqu64 %zmm21, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm0, %zmm16
+; AVX512F-NEXT:    vmovdqu64 %zmm16, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 512(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 448(%rdi), %zmm11
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = <0,7,14,21,28,u,u,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm20
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm3, %zmm20
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = <1,8,15,22,29,u,u,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm21
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm4, %zmm21
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm5 = <18,25,0,7,14,u,u,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm22
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm5, %zmm22
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm8 = <19,26,1,8,15,u,u,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm23
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm8, %zmm23
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} xmm18 = [4,11,18,25]
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm18, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} xmm19 = [5,12,19,26]
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm19, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,13,20,27]
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm1, %zmm11
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2
+; AVX512F-NEXT:    vmovdqa64 (%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 64(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm15
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm3, %zmm15
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm14
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm4, %zmm14
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm16
+; AVX512F-NEXT:    vpermt2d %zmm6, %zmm5, %zmm16
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm17
+; AVX512F-NEXT:    vpermt2d %zmm6, %zmm8, %zmm17
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm30
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm18, %zmm30
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm19, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm2, %zmm6
+; AVX512F-NEXT:    vmovdqa64 960(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa64 896(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm9
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm3, %zmm9
+; AVX512F-NEXT:    vmovdqa64 1408(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 1344(%rdi), %zmm2
+; AVX512F-NEXT:    vpermi2d %zmm0, %zmm2, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm10
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm4, %zmm10
+; AVX512F-NEXT:    vpermi2d %zmm0, %zmm2, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm12
+; AVX512F-NEXT:    vpermt2d %zmm7, %zmm5, %zmm12
+; AVX512F-NEXT:    vpermi2d %zmm2, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm13
+; AVX512F-NEXT:    vpermt2d %zmm7, %zmm8, %zmm13
+; AVX512F-NEXT:    vpermi2d %zmm2, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm25
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm18, %zmm25
+; AVX512F-NEXT:    vpermi2d %zmm0, %zmm2, %zmm18
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm26
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm19, %zmm26
+; AVX512F-NEXT:    vpermi2d %zmm0, %zmm2, %zmm19
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} xmm31 = [6,13,20,27]
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm31, %zmm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm0 = [6,13,20,27]
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm0, %zmm7
+; AVX512F-NEXT:    movw $992, %ax # imm = 0x3E0
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm15 {%k1}
+; AVX512F-NEXT:    movb $-32, %al
+; AVX512F-NEXT:    kmovw %eax, %k2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm20 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k2}
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqa32 %zmm24, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
+; AVX512F-NEXT:    movw $480, %ax # imm = 0x1E0
+; AVX512F-NEXT:    kmovw %eax, %k2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm10 {%k2}
+; AVX512F-NEXT:    movw $-512, %ax # imm = 0xFE00
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm10 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm14 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm14 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm12 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm12 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm22 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm22 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm13 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm13 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm23 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm23 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm17 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm17 {%k1}
+; AVX512F-NEXT:    vmovdqa32 %zmm27, %zmm8 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm0, %zmm8 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm25, %zmm0, %zmm0
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm1, %zmm0 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm1 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm24, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm30, %zmm24, %zmm24
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm25, %zmm24 {%k1}
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm18, %zmm28, %zmm18
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm25, %zmm18 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm26, %zmm25, %zmm25
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm26, %zmm25 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm26, %zmm26 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm27, %zmm26 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm27, %zmm27 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm28, %zmm27 {%k1}
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm19, %zmm29, %zmm19
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm28, %zmm19 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm11, %zmm28, %zmm11
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm28, %zmm11 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm6, %zmm28, %zmm6
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm28, %zmm6 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm2, %zmm28, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm28, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm7, %zmm28, %zmm7
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa32 %zmm28, %zmm7 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm3, 192(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm9, 128(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm20, 64(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm15, (%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm4, 192(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm14, (%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm21, 64(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm10, 128(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm5, 192(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm16, (%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm22, 64(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm12, 128(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm8, 192(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm17, (%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm23, 64(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm13, 128(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm18, 192(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm24, (%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm1, 64(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm0, 128(%r9)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovdqa64 %zmm19, 192(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm27, (%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm26, 64(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm25, 128(%rax)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovdqa64 %zmm7, 128(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm2, 192(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm6, (%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm11, 64(%rax)
+; AVX512F-NEXT:    addq $3400, %rsp # imm = 0xD48
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: load_i32_stride7_vf64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    subq $3400, %rsp # imm = 0xD48
+; AVX512BW-NEXT:    vmovdqa64 1728(%rdi), %zmm13
+; AVX512BW-NEXT:    vmovdqa64 1664(%rdi), %zmm14
+; AVX512BW-NEXT:    vmovdqa64 1600(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqa64 1280(%rdi), %zmm10
+; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %zmm12
+; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %zmm5
+; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm17
+; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm18
+; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm15
+; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm8
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [20,27,6,13,20,27,6,13,20,27,6,13,20,27,6,13]
+; AVX512BW-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm2, %zmm3
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,10,11,12,13,18,25,0,0,10,11,12,13,18,25]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm12, %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm14, %zmm4, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,7,14,21,28,0,0,0,0,7,14,21,28,0,0]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm12, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,19,26,0,9,10,11,12,13,19,26]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm14, %zmm4, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,1,8,15,22,29,0,0,0,1,8,15,22,29,0,0]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm12, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,20,27,0,9,10,11,12,13,20,27]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm14, %zmm4, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,18,25,0,7,14,0,0,0,18,25,0,7,14,0,0]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm5, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,21,28,0,9,10,11,12,13,21,28]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm8, %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm4, %zmm14, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,19,26,1,8,15,0,0,0,19,26,1,8,15,0,0]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm5, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,22,29,0,9,10,11,12,13,22,29]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm8, %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm4, %zmm14, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %zmm11
+; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %zmm16
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm22 = [25,4,11,18,25,4,11,18,25,4,11,18,25,4,11,18]
+; AVX512BW-NEXT:    # zmm22 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm16, %zmm22, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm19
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm22, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm20
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm22, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1472(%rdi), %zmm9
+; AVX512BW-NEXT:    vmovdqa64 1536(%rdi), %zmm21
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm21, %zmm22, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm12, %zmm22, %zmm7
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,16,23,30,0,9,10,11,12,16,23,30]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm0, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm22, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm0, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm22, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm0, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm14, %zmm4, %zmm22
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm0, %zmm22
+; AVX512BW-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [26,5,12,19,26,5,12,19,26,5,12,19,26,5,12,19]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm0, %zmm7
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,9,10,11,12,17,24,31,0,9,10,11,12,17,24,31]
+; AVX512BW-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm6, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm0, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm6, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm14, %zmm0, %zmm4
+; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm6, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm12, %zmm0, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm6, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm16, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm21, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm24 = [8,15,0,0,0,19,26,1,8,15,0,0,0,19,26,1]
+; AVX512BW-NEXT:    # zmm24 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm24, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm27 = [27,6,13,20,27,6,13,20,27,6,13,20,27,6,13,20]
+; AVX512BW-NEXT:    # zmm27 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm27, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm28 = [28,0,0,0,0,7,14,21,28,0,0,0,0,7,14,21]
+; AVX512BW-NEXT:    # zmm28 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm28, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm29 = [29,0,0,0,1,8,15,22,29,0,0,0,1,8,15,22]
+; AVX512BW-NEXT:    # zmm29 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm29, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [14,0,0,0,18,25,0,7,14,0,0,0,18,25,0,7]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm0, %zmm19
+; AVX512BW-NEXT:    vmovdqu64 %zmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm24, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm27, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm28, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm29, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm0, %zmm20
+; AVX512BW-NEXT:    vmovdqu64 %zmm20, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm24, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm9, %zmm21, %zmm24
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm16, %zmm27, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm21, %zmm9, %zmm27
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm16, %zmm28, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm21, %zmm9, %zmm28
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm16, %zmm29, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm21, %zmm9, %zmm29
+; AVX512BW-NEXT:    vpermt2d %zmm9, %zmm0, %zmm21
+; AVX512BW-NEXT:    vmovdqu64 %zmm21, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm0, %zmm16
+; AVX512BW-NEXT:    vmovdqu64 %zmm16, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm11
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm3 = <0,7,14,21,28,u,u,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm20
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm3, %zmm20
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm4 = <1,8,15,22,29,u,u,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm21
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm4, %zmm21
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm5 = <18,25,0,7,14,u,u,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm22
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm5, %zmm22
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm8 = <19,26,1,8,15,u,u,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm23
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm8, %zmm23
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} xmm18 = [4,11,18,25]
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm18, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} xmm19 = [5,12,19,26]
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm19, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,13,20,27]
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm1, %zmm11
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm15
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm3, %zmm15
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm14
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm4, %zmm14
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm16
+; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm5, %zmm16
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm17
+; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm8, %zmm17
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm30
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm18, %zmm30
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm19, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm2, %zmm6
+; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm9
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm3, %zmm9
+; AVX512BW-NEXT:    vmovdqa64 1408(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 1344(%rdi), %zmm2
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm2, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm10
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm4, %zmm10
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm2, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm12
+; AVX512BW-NEXT:    vpermt2d %zmm7, %zmm5, %zmm12
+; AVX512BW-NEXT:    vpermi2d %zmm2, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm13
+; AVX512BW-NEXT:    vpermt2d %zmm7, %zmm8, %zmm13
+; AVX512BW-NEXT:    vpermi2d %zmm2, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm25
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm18, %zmm25
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm2, %zmm18
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm26
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm19, %zmm26
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm2, %zmm19
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} xmm31 = [6,13,20,27]
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm31, %zmm2
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm0 = [6,13,20,27]
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm0, %zmm7
+; AVX512BW-NEXT:    movw $992, %ax # imm = 0x3E0
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm15 {%k1}
+; AVX512BW-NEXT:    movb $-32, %al
+; AVX512BW-NEXT:    kmovd %eax, %k2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm20 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqa32 %zmm24, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
+; AVX512BW-NEXT:    movw $480, %ax # imm = 0x1E0
+; AVX512BW-NEXT:    kmovd %eax, %k2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm10 {%k2}
+; AVX512BW-NEXT:    movw $-512, %ax # imm = 0xFE00
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm10 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm14 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm14 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm12 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm12 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm22 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm22 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm13 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm13 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm23 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm23 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm17 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm17 {%k1}
+; AVX512BW-NEXT:    vmovdqa32 %zmm27, %zmm8 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm8 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm25, %zmm0, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm1, %zmm0 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm1 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm24, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm30, %zmm24, %zmm24
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm25, %zmm24 {%k1}
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm18, %zmm28, %zmm18
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm25, %zmm18 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm26, %zmm25, %zmm25
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm26, %zmm25 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm26, %zmm26 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm27, %zmm26 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm27, %zmm27 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm28, %zmm27 {%k1}
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm19, %zmm29, %zmm19
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm28, %zmm19 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm11, %zmm28, %zmm11
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm28, %zmm11 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm6, %zmm28, %zmm6
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm28, %zmm6 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm2, %zmm28, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm28, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm7, %zmm28, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa32 %zmm28, %zmm7 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, 192(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, 128(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, 64(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, (%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, 192(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, (%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, 64(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, 128(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, 192(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, (%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, 64(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, 128(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, 192(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, (%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, 64(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, 128(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, 192(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, (%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, 64(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, 128(%r9)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, 192(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, (%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, 64(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, 128(%rax)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, 128(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, 192(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, (%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, 64(%rax)
+; AVX512BW-NEXT:    addq $3400, %rsp # imm = 0xD48
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %wide.vec = load <448 x i32>, ptr %in.vec, align 64
+  %strided.vec0 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49, i32 56, i32 63, i32 70, i32 77, i32 84, i32 91, i32 98, i32 105, i32 112, i32 119, i32 126, i32 133, i32 140, i32 147, i32 154, i32 161, i32 168, i32 175, i32 182, i32 189, i32 196, i32 203, i32 210, i32 217, i32 224, i32 231, i32 238, i32 245, i32 252, i32 259, i32 266, i32 273, i32 280, i32 287, i32 294, i32 301, i32 308, i32 315, i32 322, i32 329, i32 336, i32 343, i32 350, i32 357, i32 364, i32 371, i32 378, i32 385, i32 392, i32 399, i32 406, i32 413, i32 420, i32 427, i32 434, i32 441>
+  %strided.vec1 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50, i32 57, i32 64, i32 71, i32 78, i32 85, i32 92, i32 99, i32 106, i32 113, i32 120, i32 127, i32 134, i32 141, i32 148, i32 155, i32 162, i32 169, i32 176, i32 183, i32 190, i32 197, i32 204, i32 211, i32 218, i32 225, i32 232, i32 239, i32 246, i32 253, i32 260, i32 267, i32 274, i32 281, i32 288, i32 295, i32 302, i32 309, i32 316, i32 323, i32 330, i32 337, i32 344, i32 351, i32 358, i32 365, i32 372, i32 379, i32 386, i32 393, i32 400, i32 407, i32 414, i32 421, i32 428, i32 435, i32 442>
+  %strided.vec2 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51, i32 58, i32 65, i32 72, i32 79, i32 86, i32 93, i32 100, i32 107, i32 114, i32 121, i32 128, i32 135, i32 142, i32 149, i32 156, i32 163, i32 170, i32 177, i32 184, i32 191, i32 198, i32 205, i32 212, i32 219, i32 226, i32 233, i32 240, i32 247, i32 254, i32 261, i32 268, i32 275, i32 282, i32 289, i32 296, i32 303, i32 310, i32 317, i32 324, i32 331, i32 338, i32 345, i32 352, i32 359, i32 366, i32 373, i32 380, i32 387, i32 394, i32 401, i32 408, i32 415, i32 422, i32 429, i32 436, i32 443>
+  %strided.vec3 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> <i32 3, i32 10, i32 17, i32 24, i32 31, i32 38, i32 45, i32 52, i32 59, i32 66, i32 73, i32 80, i32 87, i32 94, i32 101, i32 108, i32 115, i32 122, i32 129, i32 136, i32 143, i32 150, i32 157, i32 164, i32 171, i32 178, i32 185, i32 192, i32 199, i32 206, i32 213, i32 220, i32 227, i32 234, i32 241, i32 248, i32 255, i32 262, i32 269, i32 276, i32 283, i32 290, i32 297, i32 304, i32 311, i32 318, i32 325, i32 332, i32 339, i32 346, i32 353, i32 360, i32 367, i32 374, i32 381, i32 388, i32 395, i32 402, i32 409, i32 416, i32 423, i32 430, i32 437, i32 444>
+  %strided.vec4 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53, i32 60, i32 67, i32 74, i32 81, i32 88, i32 95, i32 102, i32 109, i32 116, i32 123, i32 130, i32 137, i32 144, i32 151, i32 158, i32 165, i32 172, i32 179, i32 186, i32 193, i32 200, i32 207, i32 214, i32 221, i32 228, i32 235, i32 242, i32 249, i32 256, i32 263, i32 270, i32 277, i32 284, i32 291, i32 298, i32 305, i32 312, i32 319, i32 326, i32 333, i32 340, i32 347, i32 354, i32 361, i32 368, i32 375, i32 382, i32 389, i32 396, i32 403, i32 410, i32 417, i32 424, i32 431, i32 438, i32 445>
+  %strided.vec5 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54, i32 61, i32 68, i32 75, i32 82, i32 89, i32 96, i32 103, i32 110, i32 117, i32 124, i32 131, i32 138, i32 145, i32 152, i32 159, i32 166, i32 173, i32 180, i32 187, i32 194, i32 201, i32 208, i32 215, i32 222, i32 229, i32 236, i32 243, i32 250, i32 257, i32 264, i32 271, i32 278, i32 285, i32 292, i32 299, i32 306, i32 313, i32 320, i32 327, i32 334, i32 341, i32 348, i32 355, i32 362, i32 369, i32 376, i32 383, i32 390, i32 397, i32 404, i32 411, i32 418, i32 425, i32 432, i32 439, i32 446>
+  %strided.vec6 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55, i32 62, i32 69, i32 76, i32 83, i32 90, i32 97, i32 104, i32 111, i32 118, i32 125, i32 132, i32 139, i32 146, i32 153, i32 160, i32 167, i32 174, i32 181, i32 188, i32 195, i32 202, i32 209, i32 216, i32 223, i32 230, i32 237, i32 244, i32 251, i32 258, i32 265, i32 272, i32 279, i32 286, i32 293, i32 300, i32 307, i32 314, i32 321, i32 328, i32 335, i32 342, i32 349, i32 356, i32 363, i32 370, i32 377, i32 384, i32 391, i32 398, i32 405, i32 412, i32 419, i32 426, i32 433, i32 440, i32 447>
+  store <64 x i32> %strided.vec0, ptr %out.vec0, align 64
+  store <64 x i32> %strided.vec1, ptr %out.vec1, align 64
+  store <64 x i32> %strided.vec2, ptr %out.vec2, align 64
+  store <64 x i32> %strided.vec3, ptr %out.vec3, align 64
+  store <64 x i32> %strided.vec4, ptr %out.vec4, align 64
+  store <64 x i32> %strided.vec5, ptr %out.vec5, align 64
+  store <64 x i32> %strided.vec6, ptr %out.vec6, align 64
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
index 7b6c8f6..3d013fd 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
@@ -247,7 +247,6 @@ define void @load_i32_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <16 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <16 x i32> %wide.vec, <16 x i32> poison, <2 x i32> <i32 0, i32 8>
   %strided.vec1 = shufflevector <16 x i32> %wide.vec, <16 x i32> poison, <2 x i32> <i32 1, i32 9>
   %strided.vec2 = shufflevector <16 x i32> %wide.vec, <16 x i32> poison, <2 x i32> <i32 2, i32 10>
@@ -256,7 +255,6 @@ define void @load_i32_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <16 x i32> %wide.vec, <16 x i32> poison, <2 x i32> <i32 5, i32 13>
   %strided.vec6 = shufflevector <16 x i32> %wide.vec, <16 x i32> poison, <2 x i32> <i32 6, i32 14>
   %strided.vec7 = shufflevector <16 x i32> %wide.vec, <16 x i32> poison, <2 x i32> <i32 7, i32 15>
-
   store <2 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i32> %strided.vec2, ptr %out.vec2, align 64
@@ -265,7 +263,6 @@ define void @load_i32_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <2 x i32> %strided.vec5, ptr %out.vec5, align 64
   store <2 x i32> %strided.vec6, ptr %out.vec6, align 64
   store <2 x i32> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -946,7 +943,6 @@ define void @load_i32_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512DQBW-FAST-NEXT:    popq %rbx
 ; AVX512DQBW-FAST-NEXT:    retq
   %wide.vec = load <32 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <32 x i32> %wide.vec, <32 x i32> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
   %strided.vec1 = shufflevector <32 x i32> %wide.vec, <32 x i32> poison, <4 x i32> <i32 1, i32 9, i32 17, i32 25>
   %strided.vec2 = shufflevector <32 x i32> %wide.vec, <32 x i32> poison, <4 x i32> <i32 2, i32 10, i32 18, i32 26>
@@ -955,7 +951,6 @@ define void @load_i32_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <32 x i32> %wide.vec, <32 x i32> poison, <4 x i32> <i32 5, i32 13, i32 21, i32 29>
   %strided.vec6 = shufflevector <32 x i32> %wide.vec, <32 x i32> poison, <4 x i32> <i32 6, i32 14, i32 22, i32 30>
   %strided.vec7 = shufflevector <32 x i32> %wide.vec, <32 x i32> poison, <4 x i32> <i32 7, i32 15, i32 23, i32 31>
-
   store <4 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i32> %strided.vec2, ptr %out.vec2, align 64
@@ -964,7 +959,6 @@ define void @load_i32_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <4 x i32> %strided.vec5, ptr %out.vec5, align 64
   store <4 x i32> %strided.vec6, ptr %out.vec6, align 64
   store <4 x i32> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -1355,7 +1349,6 @@ define void @load_i32_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <64 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <64 x i32> %wide.vec, <64 x i32> poison, <8 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56>
   %strided.vec1 = shufflevector <64 x i32> %wide.vec, <64 x i32> poison, <8 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57>
   %strided.vec2 = shufflevector <64 x i32> %wide.vec, <64 x i32> poison, <8 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58>
@@ -1364,7 +1357,6 @@ define void @load_i32_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <64 x i32> %wide.vec, <64 x i32> poison, <8 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61>
   %strided.vec6 = shufflevector <64 x i32> %wide.vec, <64 x i32> poison, <8 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62>
   %strided.vec7 = shufflevector <64 x i32> %wide.vec, <64 x i32> poison, <8 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63>
-
   store <8 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i32> %strided.vec2, ptr %out.vec2, align 64
@@ -1373,7 +1365,6 @@ define void @load_i32_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <8 x i32> %strided.vec5, ptr %out.vec5, align 64
   store <8 x i32> %strided.vec6, ptr %out.vec6, align 64
   store <8 x i32> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -2400,7 +2391,6 @@ define void @load_i32_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <128 x i32>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <16 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120>
   %strided.vec1 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <16 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121>
   %strided.vec2 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <16 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122>
@@ -2409,7 +2399,6 @@ define void @load_i32_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   %strided.vec5 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <16 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125>
   %strided.vec6 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <16 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126>
   %strided.vec7 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <16 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127>
-
   store <16 x i32> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i32> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i32> %strided.vec2, ptr %out.vec2, align 64
@@ -2418,7 +2407,7390 @@ define void @load_i32_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   store <16 x i32> %strided.vec5, ptr %out.vec5, align 64
   store <16 x i32> %strided.vec6, ptr %out.vec6, align 64
   store <16 x i32> %strided.vec7, ptr %out.vec7, align 64
+  ret void
+}
 
+define void @load_i32_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
+; SSE-LABEL: load_i32_stride8_vf32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    subq $952, %rsp # imm = 0x3B8
+; SSE-NEXT:    movaps 544(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 608(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 576(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 672(%rdi), %xmm6
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 640(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 736(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 704(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 160(%rdi), %xmm11
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 128(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 224(%rdi), %xmm12
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 192(%rdi), %xmm8
+; SSE-NEXT:    movaps %xmm8, %xmm9
+; SSE-NEXT:    unpcklps {{.*#+}} xmm9 = xmm9[0],xmm12[0],xmm9[1],xmm12[1]
+; SSE-NEXT:    movaps %xmm0, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
+; SSE-NEXT:    movaps %xmm1, %xmm0
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm9[0]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm10 = xmm10[0],xmm7[0],xmm10[1],xmm7[1]
+; SSE-NEXT:    movaps %xmm2, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm10[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT:    movaps 512(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 480(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 448(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 416(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 384(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 992(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 960(%rdi), %xmm9
+; SSE-NEXT:    movaps %xmm9, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 928(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 896(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 352(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 320(%rdi), %xmm12
+; SSE-NEXT:    movaps %xmm12, %xmm11
+; SSE-NEXT:    unpcklps {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT:    movaps 288(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 256(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm11[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm11[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 864(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps 832(%rdi), %xmm11
+; SSE-NEXT:    movaps %xmm11, %xmm13
+; SSE-NEXT:    unpcklps {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
+; SSE-NEXT:    movaps 800(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 768(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm13[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm13[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 96(%rdi), %xmm15
+; SSE-NEXT:    movaps 64(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, %xmm13
+; SSE-NEXT:    unpcklps {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1]
+; SSE-NEXT:    movaps (%rdi), %xmm6
+; SSE-NEXT:    movaps 32(%rdi), %xmm14
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm14[0],xmm5[1],xmm14[1]
+; SSE-NEXT:    movaps %xmm5, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm13[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm13[1]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = xmm13[2],mem[2],xmm13[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[2],mem[2],xmm9[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    unpckhps (%rsp), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = xmm11[2],mem[2],xmm11[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm14[2],xmm6[3],xmm14[3]
+; SSE-NEXT:    movaps %xmm4, %xmm5
+; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm8[0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm8[1]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm1, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm12[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm12[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm10[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm13[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm13[1]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, %xmm3
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm4[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm11[0]
+; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm11[1]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm7, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm9[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 240(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 208(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movaps 176(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 144(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 368(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 336(%rdi), %xmm9
+; SSE-NEXT:    movaps %xmm9, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 304(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 272(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 496(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 464(%rdi), %xmm8
+; SSE-NEXT:    movaps %xmm8, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 432(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 400(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 624(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 592(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 560(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 528(%rdi), %xmm13
+; SSE-NEXT:    movaps %xmm13, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 752(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 720(%rdi), %xmm6
+; SSE-NEXT:    movaps %xmm6, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 688(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 656(%rdi), %xmm11
+; SSE-NEXT:    movaps %xmm11, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 880(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 848(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 816(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 784(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1008(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 976(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 944(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 912(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, %xmm15
+; SSE-NEXT:    unpcklps {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm15, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm0[1]
+; SSE-NEXT:    movaps 112(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 16(%rdi), %xmm12
+; SSE-NEXT:    movaps 48(%rdi), %xmm14
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm12, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
+; SSE-NEXT:    movaps %xmm1, %xmm14
+; SSE-NEXT:    movlhps {{.*#+}} xmm14 = xmm14[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = xmm14[2],mem[2],xmm14[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[2],mem[2],xmm9[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = xmm13[2],mem[2],xmm13[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = xmm11[2],mem[2],xmm11[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps %xmm1, %xmm0
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm14[0]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm14[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm9[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm9[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm14, %xmm9
+; SSE-NEXT:    movlhps {{.*#+}} xmm9 = xmm9[0],xmm8[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm8[1]
+; SSE-NEXT:    movaps %xmm13, %xmm8
+; SSE-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm7[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm7[1]
+; SSE-NEXT:    movaps %xmm11, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm6[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm6[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, %xmm0
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm5[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm5[1]
+; SSE-NEXT:    movaps %xmm1, %xmm6
+; SSE-NEXT:    movaps %xmm10, %xmm5
+; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm12, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm2[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm2[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rdx)
+; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%r9)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm15, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rax)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm5, 112(%rax)
+; SSE-NEXT:    movaps %xmm0, 96(%rax)
+; SSE-NEXT:    movaps %xmm7, 80(%rax)
+; SSE-NEXT:    movaps %xmm8, 64(%rax)
+; SSE-NEXT:    movaps %xmm9, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps %xmm4, (%rax)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm10, 112(%rax)
+; SSE-NEXT:    movaps %xmm6, 96(%rax)
+; SSE-NEXT:    movaps %xmm11, 80(%rax)
+; SSE-NEXT:    movaps %xmm13, 64(%rax)
+; SSE-NEXT:    movaps %xmm14, 48(%rax)
+; SSE-NEXT:    movaps %xmm3, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps %xmm12, (%rax)
+; SSE-NEXT:    addq $952, %rsp # imm = 0x3B8
+; SSE-NEXT:    retq
+;
+; AVX1-ONLY-LABEL: load_i32_stride8_vf32:
+; AVX1-ONLY:       # %bb.0:
+; AVX1-ONLY-NEXT:    subq $1736, %rsp # imm = 0x6C8
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm10[0]
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm15[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm4[0],xmm11[0],xmm4[1],xmm11[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm7[0],xmm13[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm12[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3,4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm5[0],xmm7[0],xmm5[1],xmm7[1]
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm1[0],xmm3[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm9[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm10[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm10[1,1,1,1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],xmm8[1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm6[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm6 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm6 = xmm6[0],mem[1],xmm6[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm4[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm11[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0],xmm6[1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, %xmm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm7[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps $2, (%rsp), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm2[0],mem[1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm14[2],xmm9[2],xmm14[3],xmm9[3]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm10[2],xmm8[2],xmm10[3],xmm8[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = mem[0,1,2],xmm2[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm15[0,1,2],xmm3[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm1[0,1],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = mem[0,1,2],xmm2[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm10[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm2[2],xmm6[2],xmm2[3],xmm6[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm13[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm12[0,1,2],xmm2[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm5[0,1,2],xmm3[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm11[2],xmm7[2],xmm11[3],xmm7[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps (%rsp), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm11[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm9[0,1,2],xmm7[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm3[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm2[0,1,2],xmm8[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm7 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm7[1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm4[2],xmm10[2],xmm4[3],xmm10[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm12[2],xmm13[2],xmm12[3],xmm13[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm14[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm9[2],xmm11[2],xmm9[3],xmm11[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps $238, (%rsp), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[2],ymm6[2]
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm3[0],ymm8[0],ymm3[1],ymm8[1],ymm3[4],ymm8[4],ymm3[5],ymm8[5]
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[1],ymm7[1],ymm1[4],ymm7[4],ymm1[5],ymm7[5]
+; AVX1-ONLY-NEXT:    vmovups %ymm7, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, %ymm14
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
+; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %ymm11
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm0[0],ymm11[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm3[0],ymm1[0],ymm3[1],ymm1[1],ymm3[4],ymm1[4],ymm3[5],ymm1[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm12
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm5[0],ymm0[0],ymm5[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm12[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm13 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[0,1],ymm12[2,0],ymm13[4,5],ymm12[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm12, %xmm0
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm15
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm13 = ymm2[0],ymm15[0],ymm2[1],ymm15[1],ymm2[4],ymm15[4],ymm2[5],ymm15[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm13[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm6[0],ymm3[0],ymm6[1],ymm3[1],ymm6[4],ymm3[4],ymm6[5],ymm3[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,0],ymm12[1,0],ymm8[5,4],ymm12[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm14[0],ymm1[1],ymm14[1],ymm1[4],ymm14[4],ymm1[5],ymm14[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm7[1,0],ymm6[1,0],ymm7[5,4],ymm6[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm9[0],ymm10[0],ymm9[1],ymm10[1],ymm9[4],ymm10[4],ymm9[5],ymm10[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm14[1,0],ymm8[1,0],ymm14[5,4],ymm8[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm4[0],mem[0],ymm4[1],mem[1],ymm4[4],mem[4],ymm4[5],mem[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm7[1,0],ymm9[1,0],ymm7[5,4],ymm9[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm11[0],ymm0[1],ymm11[1],ymm0[4],ymm11[4],ymm0[5],ymm11[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[4],ymm5[4],ymm4[5],ymm5[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm5[1,0],ymm2[1,0],ymm5[5,4],ymm2[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm11[0],ymm0[1],ymm11[1],ymm0[4],ymm11[4],ymm0[5],ymm11[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[4],mem[4],ymm1[5],mem[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm13 = ymm15[1,0],mem[1,0],ymm15[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm3[1],mem[1],ymm3[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vunpckhps (%rsp), %ymm6, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm6[2],mem[2],ymm6[3],mem[3],ymm6[6],mem[6],ymm6[7],mem[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm3[1],ymm12[1],ymm3[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm10[1],mem[1],ymm10[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm14[2],ymm8[3],ymm14[3],ymm8[6],ymm14[6],ymm8[7],ymm14[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm7[2],ymm9[3],ymm7[3],ymm9[6],ymm7[6],ymm9[7],ymm7[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm13 = ymm3[1],mem[1],ymm3[3],mem[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm6[1],ymm8[1],ymm6[3],ymm8[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm7[2],ymm3[3],ymm7[3],ymm3[6],ymm7[6],ymm3[7],ymm7[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm5[2],ymm2[3],ymm5[3],ymm2[6],ymm5[6],ymm2[7],ymm5[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm2[1],ymm4[1],ymm2[3],ymm4[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm14[1],ymm11[3],ymm14[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm10[2],ymm4[3],ymm10[3],ymm4[6],ymm10[6],ymm4[7],ymm10[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm13 = ymm9[2],ymm15[2],ymm9[3],ymm15[3],ymm9[6],ymm15[6],ymm9[7],ymm15[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm11 = ymm11[3,0],mem[3,0],ymm11[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm11, %xmm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm11[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm8[2],ymm6[2],ymm8[3],ymm6[3],ymm8[6],ymm6[6],ymm8[7],ymm6[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[3,0],ymm3[3,0],ymm7[7,4],ymm3[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm2[3,0],mem[3,0],ymm2[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm2[3,0],mem[3,0],ymm2[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm1[2,3],ymm2[6,4],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm2[2],mem[2],ymm2[3],mem[3],ymm2[6],mem[6],ymm2[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[3,0],mem[3,0],ymm3[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm14[2],mem[2],ymm14[3],mem[3],ymm14[6],mem[6],ymm14[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm10[3,0],ymm4[3,0],ymm10[7,4],ymm4[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[2,0],ymm2[2,3],ymm3[6,4],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm3 = ymm3[2],ymm5[2],ymm3[3],ymm5[3],ymm3[6],ymm5[6],ymm3[7],ymm5[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm15[3,0],ymm9[3,0],ymm15[7,4],ymm9[7,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%r9)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%rax)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm13, (%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%rax)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm11, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%rax)
+; AVX1-ONLY-NEXT:    addq $1736, %rsp # imm = 0x6C8
+; AVX1-ONLY-NEXT:    vzeroupper
+; AVX1-ONLY-NEXT:    retq
+;
+; AVX2-ONLY-LABEL: load_i32_stride8_vf32:
+; AVX2-ONLY:       # %bb.0:
+; AVX2-ONLY-NEXT:    subq $1256, %rsp # imm = 0x4E8
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, %xmm11
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm2, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, %xmm13
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm3, %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, %xmm10
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm3, %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, %xmm15
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm3, %xmm3
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 800(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 768(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps 864(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps 832(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm12, %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 992(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps 960(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm2, %xmm2
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps 928(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 896(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 608(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm0, %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 576(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps 544(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 512(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 736(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps 704(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm2, %xmm2
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps 672(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 640(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm0, %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm8, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm7[0],xmm9[0],xmm7[1],xmm9[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm6, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm5, %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm14
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm14[0],xmm4[0],xmm14[1],xmm4[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm11[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm10[0],xmm13[0],xmm10[1],xmm13[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm13[0],xmm15[0],xmm13[1],xmm15[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vmovaps (%rsp), %xmm11 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm11[1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm12[0],xmm10[0],xmm12[1],xmm10[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm2[0],mem[1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm4[1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm8[0],mem[0],xmm8[1],mem[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm7[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm9[1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm8[1,1,1,1]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm7[1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm6[1,1,1,1]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm5[1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm14 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm14 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm0[2],xmm11[2],xmm0[3],xmm11[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm15[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm13[0,1,2],xmm0[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm15
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm1[0,1,2],xmm15[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm14[0,1],xmm15[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm11[2],mem[2],xmm11[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm12[2,2,2,2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm12[0,1,2],xmm14[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm13
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm10[2,2,2,2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm0[0,1,2],xmm14[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm13 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
+; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm3[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm4[0,1,2],xmm8[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm14
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm14[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm9[2,2,2,2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm3[0,1,2],xmm14[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm13[0,1],xmm14[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm14[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm9 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm9 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm14[2,2,2,2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm5[0,1,2],xmm6[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm8[2,2,2,2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm11[0,1,2],xmm4[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm4 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[2,3,2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm15[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm12[2],mem[2],xmm12[3],mem[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps $238, (%rsp), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[2,3,2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm13[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm10[2,3,2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm5[2],xmm14[2],xmm5[3],xmm14[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm9[2,3,2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm11[2],xmm8[2],xmm11[3],xmm8[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 304(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpbroadcastd 272(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovdqa 352(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 320(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovdqa 480(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 448(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 416(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovdqa 384(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm9[0],ymm1[1],ymm9[1],ymm1[4],ymm9[4],ymm1[5],ymm9[5]
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm3 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
+; AVX2-ONLY-NEXT:    vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 816(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpbroadcastd 784(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovdqa 864(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 832(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovdqa 992(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 960(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 928(%rdi), %ymm10
+; AVX2-ONLY-NEXT:    vmovdqa 896(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm2 = ymm2[0],ymm10[0],ymm2[1],ymm10[1],ymm2[4],ymm10[4],ymm2[5],ymm10[5]
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm4[0],ymm1[0],ymm4[1],ymm1[1],ymm4[4],ymm1[4],ymm4[5],ymm1[5]
+; AVX2-ONLY-NEXT:    vpunpcklqdq {{.*#+}} ymm2 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 48(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpbroadcastd 16(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovdqa 96(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 64(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vmovdqa 224(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 192(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 160(%rdi), %ymm12
+; AVX2-ONLY-NEXT:    vmovdqa 128(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm4 = ymm4[0],ymm12[0],ymm4[1],ymm12[1],ymm4[4],ymm12[4],ymm4[5],ymm12[5]
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm2 = ymm5[0],ymm2[0],ymm5[1],ymm2[1],ymm5[4],ymm2[4],ymm5[5],ymm2[5]
+; AVX2-ONLY-NEXT:    vpunpcklqdq {{.*#+}} ymm4 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 560(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpbroadcastd 528(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovdqa 608(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 576(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm4 = ymm5[0],ymm4[0],ymm5[1],ymm4[1],ymm5[4],ymm4[4],ymm5[5],ymm4[5]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm0[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT:    vmovdqa 736(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 704(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vmovdqa 672(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovdqa 640(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm5 = ymm5[0],ymm8[0],ymm5[1],ymm8[1],ymm5[4],ymm8[4],ymm5[5],ymm8[5]
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm15[0],ymm0[0],ymm15[1],ymm0[1],ymm15[4],ymm0[4],ymm15[5],ymm0[5]
+; AVX2-ONLY-NEXT:    vpunpcklqdq {{.*#+}} ymm5 = ymm5[0],ymm0[0],ymm5[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 404(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm9[5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm4 = mem[0],zero,mem[1],zero
+; AVX2-ONLY-NEXT:    vpbroadcastd 372(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0,1,2],xmm5[3]
+; AVX2-ONLY-NEXT:    vmovdqa 288(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vmovdqa 256(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm5 = ymm6[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2,3,4],ymm7[5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 916(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm10[5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm3 = mem[0],zero,mem[1],zero
+; AVX2-ONLY-NEXT:    vpbroadcastd 884(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3]
+; AVX2-ONLY-NEXT:    vmovdqa 800(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqa 768(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm11 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm11 = ymm11[0],ymm5[1],ymm11[2,3,4],ymm5[5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm11, %xmm11
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm11[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 148(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm12[5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm2 = mem[0],zero,mem[1],zero
+; AVX2-ONLY-NEXT:    vpbroadcastd 116(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm11 = xmm2[0,1,2],xmm3[3]
+; AVX2-ONLY-NEXT:    vmovdqa 32(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqa (%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm13 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm13[0],ymm3[1],ymm13[2,3,4],ymm3[5],ymm13[6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm13, %xmm13
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm11 = xmm13[0,1],xmm11[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm11[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 660(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm8[5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm11 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm0 = mem[0],zero,mem[1],zero
+; AVX2-ONLY-NEXT:    vpbroadcastd 628(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm13 = xmm0[0,1,2],xmm1[3]
+; AVX2-ONLY-NEXT:    vmovdqa 544(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqa 512(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm14 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm14 = ymm14[0],ymm1[1],ymm14[2,3,4],ymm1[5],ymm14[6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm14, %xmm14
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm13 = xmm14[0,1],xmm13[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm11 = ymm13[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 504(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1,2,3,4,5,6],ymm11[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm11 = ymm11[2],ymm9[2],ymm11[3],ymm9[3],ymm11[6],ymm9[6],ymm11[7],ymm9[7]
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm6 = ymm6[2],ymm7[2],ymm6[3],ymm7[3],ymm6[6],ymm7[6],ymm6[7],ymm7[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm13 = ymm7[2],mem[2],ymm7[3],mem[3],ymm7[6],mem[6],ymm7[7],mem[7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm6, %xmm7
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm9 = ymm13[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1016(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm7 = ymm7[2],ymm10[2],ymm7[3],ymm10[3],ymm7[6],ymm10[6],ymm7[7],ymm10[7]
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm9 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm10 = ymm4[2],mem[2],ymm4[3],mem[3],ymm4[6],mem[6],ymm4[7],mem[7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm9, %xmm4
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm5 = ymm10[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm7[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 248(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = mem[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm5[2],ymm12[2],ymm5[3],ymm12[3],ymm5[6],ymm12[6],ymm5[7],ymm12[7]
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm12 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = ymm2[2],mem[2],ymm2[3],mem[3],ymm2[6],mem[6],ymm2[7],mem[7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm12, %xmm2
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm3 = ymm14[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 760(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm8 = ymm4[2],ymm8[2],ymm4[3],ymm8[3],ymm4[6],ymm8[6],ymm4[7],ymm8[7]
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm1 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm1, %xmm0
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm4 = ymm15[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 220(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm5[1],ymm0[1],ymm5[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm14, %xmm4
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm5 = ymm12[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 476(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm11[1],ymm0[1],ymm11[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm13, %xmm5
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm6 = ymm6[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm6, %xmm6
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 732(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm5 = ymm5[2],mem[2],ymm5[3],mem[3],ymm5[6],mem[6],ymm5[7],mem[7]
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm5 = ymm8[1],ymm5[1],ymm8[3],ymm5[3]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm15, %xmm6
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 988(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm5 = ymm5[2],mem[2],ymm5[3],mem[3],ymm5[6],mem[6],ymm5[7],mem[7]
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm5 = ymm7[1],ymm5[1],ymm7[3],ymm5[3]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm10, %xmm6
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm7 = ymm9[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm7, %xmm7
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 64(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, (%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 96(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 32(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 64(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, (%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 96(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 32(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 64(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, (%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 96(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 32(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 64(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, (%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 96(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 32(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 64(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, (%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 96(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 32(%r9)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, (%rax)
+; AVX2-ONLY-NEXT:    vmovups (%rsp), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 32(%rax)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovdqa %ymm2, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm3, (%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, 32(%rax)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovdqa %ymm5, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm1, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm0, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm4, (%rax)
+; AVX2-ONLY-NEXT:    addq $1256, %rsp # imm = 0x4E8
+; AVX2-ONLY-NEXT:    vzeroupper
+; AVX2-ONLY-NEXT:    retq
+;
+; AVX512F-LABEL: load_i32_stride8_vf32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    subq $1032, %rsp # imm = 0x408
+; AVX512F-NEXT:    vmovdqa64 704(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 512(%rdi), %zmm11
+; AVX512F-NEXT:    vmovdqa64 832(%rdi), %zmm22
+; AVX512F-NEXT:    vmovdqa64 768(%rdi), %zmm30
+; AVX512F-NEXT:    vmovdqa64 960(%rdi), %zmm29
+; AVX512F-NEXT:    vmovdqa64 896(%rdi), %zmm16
+; AVX512F-NEXT:    vmovdqa64 (%rdi), %zmm13
+; AVX512F-NEXT:    vmovdqa64 64(%rdi), %zmm31
+; AVX512F-NEXT:    vmovdqa64 128(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqa64 192(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqa64 320(%rdi), %zmm15
+; AVX512F-NEXT:    vmovdqa64 256(%rdi), %zmm28
+; AVX512F-NEXT:    vmovdqa64 448(%rdi), %zmm19
+; AVX512F-NEXT:    vmovdqa64 384(%rdi), %zmm3
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [0,8,16,24,0,8,16,24,0,8,16,24,0,8,16,24]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm10
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm0, %zmm3
+; AVX512F-NEXT:    movb $-64, %al
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,0,8,16,24>
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm7, %zmm1, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm0, %zmm5
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm4, %zmm3, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm4
+; AVX512F-NEXT:    vpermt2d %zmm22, %zmm0, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512F-NEXT:    vpermi2d %zmm9, %zmm11, %zmm0
+; AVX512F-NEXT:    vpermi2d %zmm2, %zmm6, %zmm1
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm4, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm5 = [1,9,17,25,1,9,17,25,1,9,17,25,1,9,17,25]
+; AVX512F-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm5, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = <u,u,u,u,1,9,17,25>
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm7, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm5, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [2,10,18,26,2,10,18,26,2,10,18,26,2,10,18,26]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [3,11,19,27,3,11,19,27,3,11,19,27,3,11,19,27]
+; AVX512F-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [4,12,20,28,4,12,20,28,4,12,20,28,4,12,20,28]
+; AVX512F-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm1, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm10
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm12 = [5,13,21,29,5,13,21,29,5,13,21,29,5,13,21,29]
+; AVX512F-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm12, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm14 = [6,14,22,30,6,14,22,30,6,14,22,30,6,14,22,30]
+; AVX512F-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm14, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm17 = [7,15,23,31,7,15,23,31,7,15,23,31,7,15,23,31]
+; AVX512F-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm17, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm19
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm0, %zmm19
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm4, %zmm7
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm27
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm10, %zmm27
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm6
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm12, %zmm6
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm14, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm17, %zmm28
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm1, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm9
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm4, %zmm9
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm15
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm10, %zmm15
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm12, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm14, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm17, %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm31
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm5, %zmm31
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm1, %zmm16
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm18
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm4, %zmm18
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm20
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm10, %zmm20
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm21
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm12, %zmm21
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm23
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm14, %zmm23
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm17, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm29
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm22, %zmm5, %zmm29
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm13
+; AVX512F-NEXT:    vpermt2d %zmm22, %zmm1, %zmm13
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm22
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm2, %zmm22
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm24
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm10, %zmm24
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm25
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm12, %zmm25
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm26
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm14, %zmm26
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm17, %zmm30
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm0
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2d %zmm11, %zmm0, %zmm5
+; AVX512F-NEXT:    vpermi2d %zmm11, %zmm0, %zmm1
+; AVX512F-NEXT:    vpermi2d %zmm11, %zmm0, %zmm2
+; AVX512F-NEXT:    vpermi2d %zmm11, %zmm0, %zmm4
+; AVX512F-NEXT:    vpermi2d %zmm11, %zmm0, %zmm12
+; AVX512F-NEXT:    vpermi2d %zmm11, %zmm0, %zmm14
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm17, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm29 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm0 = <u,u,u,u,1,9,17,25>
+; AVX512F-NEXT:    vpermi2d %zmm31, %zmm11, %zmm0
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm5, %zmm29, %zmm17
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm19 {%k1}
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm5 = <u,u,u,u,2,10,18,26>
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm10
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm29 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm5, %zmm10
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm10[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm3, %zmm19, %zmm19
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm13 {%k1}
+; AVX512F-NEXT:    vpermi2d %zmm31, %zmm11, %zmm5
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm5 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm5, %zmm13, %zmm16
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm7 {%k1}
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm10 = <u,u,u,u,3,11,19,27>
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm13
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm10, %zmm13
+; AVX512F-NEXT:    vmovdqa64 %zmm29, %zmm5
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm9[0,1,2,3],ymm13[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm7, %zmm29
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm22 {%k1}
+; AVX512F-NEXT:    vpermi2d %zmm31, %zmm11, %zmm10
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm0
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm10[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm7, %zmm22, %zmm22
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm9 = <u,u,u,u,4,12,20,28>
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm7
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm10
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm11
+; AVX512F-NEXT:    vpermt2d %zmm5, %zmm9, %zmm10
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm13 = <u,u,u,u,5,13,21,29>
+; AVX512F-NEXT:    vpermt2d %zmm5, %zmm13, %zmm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm5 = <u,u,u,u,6,14,22,30>
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm5, %zmm1
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} ymm18 = <u,u,u,u,7,15,23,31>
+; AVX512F-NEXT:    vpermt2d %zmm11, %zmm18, %zmm7
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm11
+; AVX512F-NEXT:    vpermi2d %zmm31, %zmm0, %zmm9
+; AVX512F-NEXT:    vpermi2d %zmm31, %zmm0, %zmm13
+; AVX512F-NEXT:    vpermi2d %zmm31, %zmm0, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm18, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm7
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm10 = ymm15[0,1,2,3],ymm10[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm27 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm10, %zmm27, %zmm2
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm9 = ymm4[0,1,2,3],ymm9[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm24 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm9, %zmm24, %zmm9
+; AVX512F-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm0 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm0 = mem[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm6 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm6, %zmm0
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm3 = ymm12[0,1,2,3],ymm13[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm25 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm3, %zmm25, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm8 {%k1}
+; AVX512F-NEXT:    vpblendd $15, (%rsp), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm8, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm26 {%k1}
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm4 = ymm14[0,1,2,3],ymm5[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm4, %zmm26, %zmm4
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm28 {%k1}
+; AVX512F-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm5 = mem[0,1,2,3],ymm11[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm5, %zmm28, %zmm5
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm30 {%k1}
+; AVX512F-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm6 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm6 = mem[0,1,2,3],ymm7[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm6, %zmm30, %zmm6
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm7, 64(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm7, (%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm17, 64(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm7, (%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm16, 64(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm19, (%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm22, 64(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm29, (%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm9, 64(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm2, (%r9)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovdqa64 %zmm3, 64(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm0, (%rax)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovdqa64 %zmm4, 64(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm1, (%rax)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovdqa64 %zmm6, 64(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm5, (%rax)
+; AVX512F-NEXT:    addq $1032, %rsp # imm = 0x408
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: load_i32_stride8_vf32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    subq $1032, %rsp # imm = 0x408
+; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm11
+; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm22
+; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm30
+; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm29
+; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm16
+; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm13
+; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm31
+; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm5
+; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm15
+; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm28
+; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm19
+; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm3
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [0,8,16,24,0,8,16,24,0,8,16,24,0,8,16,24]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm10
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm0, %zmm3
+; AVX512BW-NEXT:    movb $-64, %al
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,0,8,16,24>
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm7, %zmm1, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm0, %zmm5
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm4, %zmm3, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm4
+; AVX512BW-NEXT:    vpermt2d %zmm22, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512BW-NEXT:    vpermi2d %zmm9, %zmm11, %zmm0
+; AVX512BW-NEXT:    vpermi2d %zmm2, %zmm6, %zmm1
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm4, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm5 = [1,9,17,25,1,9,17,25,1,9,17,25,1,9,17,25]
+; AVX512BW-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm5, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = <u,u,u,u,1,9,17,25>
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm7, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm5, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [2,10,18,26,2,10,18,26,2,10,18,26,2,10,18,26]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [3,11,19,27,3,11,19,27,3,11,19,27,3,11,19,27]
+; AVX512BW-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [4,12,20,28,4,12,20,28,4,12,20,28,4,12,20,28]
+; AVX512BW-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm1, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm10
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm12 = [5,13,21,29,5,13,21,29,5,13,21,29,5,13,21,29]
+; AVX512BW-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm12, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm14 = [6,14,22,30,6,14,22,30,6,14,22,30,6,14,22,30]
+; AVX512BW-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm14, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm17 = [7,15,23,31,7,15,23,31,7,15,23,31,7,15,23,31]
+; AVX512BW-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm17, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm19
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm0, %zmm19
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm4, %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm27
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm10, %zmm27
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm6
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm12, %zmm6
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm14, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm17, %zmm28
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm1, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm9
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm4, %zmm9
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm15
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm10, %zmm15
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm12, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm14, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm17, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm31
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm5, %zmm31
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm1, %zmm16
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm18
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm4, %zmm18
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm20
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm10, %zmm20
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm21
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm12, %zmm21
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm23
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm14, %zmm23
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm17, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm29
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm22, %zmm5, %zmm29
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm13
+; AVX512BW-NEXT:    vpermt2d %zmm22, %zmm1, %zmm13
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm22
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm2, %zmm22
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm24
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm10, %zmm24
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm25
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm12, %zmm25
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm26
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm14, %zmm26
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm17, %zmm30
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2d %zmm11, %zmm0, %zmm5
+; AVX512BW-NEXT:    vpermi2d %zmm11, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpermi2d %zmm11, %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermi2d %zmm11, %zmm0, %zmm4
+; AVX512BW-NEXT:    vpermi2d %zmm11, %zmm0, %zmm12
+; AVX512BW-NEXT:    vpermi2d %zmm11, %zmm0, %zmm14
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm17, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm29 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm0 = <u,u,u,u,1,9,17,25>
+; AVX512BW-NEXT:    vpermi2d %zmm31, %zmm11, %zmm0
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm5, %zmm29, %zmm17
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm19 {%k1}
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm5 = <u,u,u,u,2,10,18,26>
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm10
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm29 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm5, %zmm10
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm10[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm3, %zmm19, %zmm19
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm13 {%k1}
+; AVX512BW-NEXT:    vpermi2d %zmm31, %zmm11, %zmm5
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm5, %zmm13, %zmm16
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm7 {%k1}
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm10 = <u,u,u,u,3,11,19,27>
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm13
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm10, %zmm13
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, %zmm5
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm9[0,1,2,3],ymm13[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm7, %zmm29
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm22 {%k1}
+; AVX512BW-NEXT:    vpermi2d %zmm31, %zmm11, %zmm10
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm0
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm10[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm7, %zmm22, %zmm22
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm9 = <u,u,u,u,4,12,20,28>
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm10
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm11
+; AVX512BW-NEXT:    vpermt2d %zmm5, %zmm9, %zmm10
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm13 = <u,u,u,u,5,13,21,29>
+; AVX512BW-NEXT:    vpermt2d %zmm5, %zmm13, %zmm3
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm5 = <u,u,u,u,6,14,22,30>
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm5, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} ymm18 = <u,u,u,u,7,15,23,31>
+; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm18, %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm11
+; AVX512BW-NEXT:    vpermi2d %zmm31, %zmm0, %zmm9
+; AVX512BW-NEXT:    vpermi2d %zmm31, %zmm0, %zmm13
+; AVX512BW-NEXT:    vpermi2d %zmm31, %zmm0, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm18, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm7
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm10 = ymm15[0,1,2,3],ymm10[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm27 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm10, %zmm27, %zmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm9 = ymm4[0,1,2,3],ymm9[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm24 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm9, %zmm24, %zmm9
+; AVX512BW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm0 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm0 = mem[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm6 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm6, %zmm0
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm12[0,1,2,3],ymm13[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm25 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm3, %zmm25, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm8 {%k1}
+; AVX512BW-NEXT:    vpblendd $15, (%rsp), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm8, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm26 {%k1}
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm14[0,1,2,3],ymm5[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm4, %zmm26, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm28 {%k1}
+; AVX512BW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm5 = mem[0,1,2,3],ymm11[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm5, %zmm28, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm30 {%k1}
+; AVX512BW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm6 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm6 = mem[0,1,2,3],ymm7[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm6, %zmm30, %zmm6
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm7, 64(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm7, (%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, 64(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm7, (%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, 64(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, (%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, 64(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, (%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, 64(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, (%r9)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, 64(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, (%rax)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, 64(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, (%rax)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, 64(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, (%rax)
+; AVX512BW-NEXT:    addq $1032, %rsp # imm = 0x408
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %wide.vec = load <256 x i32>, ptr %in.vec, align 64
+  %strided.vec0 = shufflevector <256 x i32> %wide.vec, <256 x i32> poison, <32 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120, i32 128, i32 136, i32 144, i32 152, i32 160, i32 168, i32 176, i32 184, i32 192, i32 200, i32 208, i32 216, i32 224, i32 232, i32 240, i32 248>
+  %strided.vec1 = shufflevector <256 x i32> %wide.vec, <256 x i32> poison, <32 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121, i32 129, i32 137, i32 145, i32 153, i32 161, i32 169, i32 177, i32 185, i32 193, i32 201, i32 209, i32 217, i32 225, i32 233, i32 241, i32 249>
+  %strided.vec2 = shufflevector <256 x i32> %wide.vec, <256 x i32> poison, <32 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122, i32 130, i32 138, i32 146, i32 154, i32 162, i32 170, i32 178, i32 186, i32 194, i32 202, i32 210, i32 218, i32 226, i32 234, i32 242, i32 250>
+  %strided.vec3 = shufflevector <256 x i32> %wide.vec, <256 x i32> poison, <32 x i32> <i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 51, i32 59, i32 67, i32 75, i32 83, i32 91, i32 99, i32 107, i32 115, i32 123, i32 131, i32 139, i32 147, i32 155, i32 163, i32 171, i32 179, i32 187, i32 195, i32 203, i32 211, i32 219, i32 227, i32 235, i32 243, i32 251>
+  %strided.vec4 = shufflevector <256 x i32> %wide.vec, <256 x i32> poison, <32 x i32> <i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 52, i32 60, i32 68, i32 76, i32 84, i32 92, i32 100, i32 108, i32 116, i32 124, i32 132, i32 140, i32 148, i32 156, i32 164, i32 172, i32 180, i32 188, i32 196, i32 204, i32 212, i32 220, i32 228, i32 236, i32 244, i32 252>
+  %strided.vec5 = shufflevector <256 x i32> %wide.vec, <256 x i32> poison, <32 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125, i32 133, i32 141, i32 149, i32 157, i32 165, i32 173, i32 181, i32 189, i32 197, i32 205, i32 213, i32 221, i32 229, i32 237, i32 245, i32 253>
+  %strided.vec6 = shufflevector <256 x i32> %wide.vec, <256 x i32> poison, <32 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126, i32 134, i32 142, i32 150, i32 158, i32 166, i32 174, i32 182, i32 190, i32 198, i32 206, i32 214, i32 222, i32 230, i32 238, i32 246, i32 254>
+  %strided.vec7 = shufflevector <256 x i32> %wide.vec, <256 x i32> poison, <32 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127, i32 135, i32 143, i32 151, i32 159, i32 167, i32 175, i32 183, i32 191, i32 199, i32 207, i32 215, i32 223, i32 231, i32 239, i32 247, i32 255>
+  store <32 x i32> %strided.vec0, ptr %out.vec0, align 64
+  store <32 x i32> %strided.vec1, ptr %out.vec1, align 64
+  store <32 x i32> %strided.vec2, ptr %out.vec2, align 64
+  store <32 x i32> %strided.vec3, ptr %out.vec3, align 64
+  store <32 x i32> %strided.vec4, ptr %out.vec4, align 64
+  store <32 x i32> %strided.vec5, ptr %out.vec5, align 64
+  store <32 x i32> %strided.vec6, ptr %out.vec6, align 64
+  store <32 x i32> %strided.vec7, ptr %out.vec7, align 64
+  ret void
+}
+
+define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
+; SSE-LABEL: load_i32_stride8_vf64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    subq $2232, %rsp # imm = 0x8B8
+; SSE-NEXT:    movaps 288(%rdi), %xmm6
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 352(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 320(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps 416(%rdi), %xmm8
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 384(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 480(%rdi), %xmm9
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 448(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 160(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 128(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 224(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 192(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm0
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm5, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    movaps 256(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 736(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 704(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 672(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 640(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 608(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 576(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 544(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 512(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 992(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 960(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 928(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 896(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 864(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 832(%rdi), %xmm12
+; SSE-NEXT:    movaps %xmm12, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 800(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 768(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1248(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1216(%rdi), %xmm13
+; SSE-NEXT:    movaps %xmm13, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1184(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1152(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1120(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1088(%rdi), %xmm14
+; SSE-NEXT:    movaps %xmm14, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1056(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1024(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1504(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1472(%rdi), %xmm15
+; SSE-NEXT:    movaps %xmm15, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1440(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1408(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1376(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1344(%rdi), %xmm6
+; SSE-NEXT:    movaps %xmm6, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1312(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1280(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1760(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1728(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1696(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1664(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1632(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1600(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1568(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1536(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2016(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1984(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1952(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1920(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1888(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1856(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1824(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1792(%rdi), %xmm8
+; SSE-NEXT:    movaps %xmm8, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
+; SSE-NEXT:    movaps %xmm1, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 96(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 64(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    movaps (%rdi), %xmm11
+; SSE-NEXT:    movaps 32(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm11, %xmm10
+; SSE-NEXT:    unpcklps {{.*#+}} xmm10 = xmm10[0],xmm7[0],xmm10[1],xmm7[1]
+; SSE-NEXT:    movaps %xmm10, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[2],mem[2],xmm9[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movaps (%rsp), %xmm10 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = xmm14[2],mem[2],xmm14[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = xmm13[2],mem[2],xmm13[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = xmm11[2],mem[2],xmm11[3],mem[3]
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm9[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm10[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm9[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm9[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm9[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm12[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm12[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm9[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm14[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm14[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm13[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm13[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm6[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm6
+; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm15[0]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm15[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, %xmm0
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm5[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm5[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm8, %xmm0
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm11, %xmm0
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 240(%rdi), %xmm14
+; SSE-NEXT:    movaps 208(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movaps 176(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 144(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 368(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 336(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 304(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 272(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 496(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 464(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 432(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 400(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 624(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 592(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movaps 560(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 528(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 752(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 720(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 688(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 656(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 880(%rdi), %xmm12
+; SSE-NEXT:    movaps 848(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT:    movaps 816(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 784(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1008(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 976(%rdi), %xmm15
+; SSE-NEXT:    movaps %xmm15, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 944(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 912(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1136(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1104(%rdi), %xmm9
+; SSE-NEXT:    movaps %xmm9, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1072(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1040(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1264(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1232(%rdi), %xmm8
+; SSE-NEXT:    movaps %xmm8, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1200(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1168(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1392(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1360(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1328(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1296(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1520(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1488(%rdi), %xmm6
+; SSE-NEXT:    movaps %xmm6, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1456(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1424(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1648(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1616(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1584(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1552(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1776(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1744(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1712(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1680(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1904(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1872(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1840(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1808(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2032(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2000(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    movaps 1968(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1936(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm1, %xmm10
+; SSE-NEXT:    movlhps {{.*#+}} xmm10 = xmm10[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 112(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm0
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movaps 16(%rdi), %xmm13
+; SSE-NEXT:    movaps 48(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm13, %xmm11
+; SSE-NEXT:    unpcklps {{.*#+}} xmm11 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm11, %xmm10
+; SSE-NEXT:    movlhps {{.*#+}} xmm10 = xmm10[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm14[2],xmm0[3],xmm14[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = xmm14[2],mem[2],xmm14[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = xmm11[2],mem[2],xmm11[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{.*#+}} xmm10 = xmm10[2],xmm12[2],xmm10[3],xmm12[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[2],mem[2],xmm9[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; SSE-NEXT:    movaps (%rsp), %xmm12 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps %xmm12, (%rsp) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = xmm13[2],mem[2],xmm13[3],mem[3]
+; SSE-NEXT:    movaps %xmm14, %xmm12
+; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm14, %xmm12
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm14, %xmm12
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm14, %xmm12
+; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm11[0]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm11[1]
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, %xmm12
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm11[0]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm11[1]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm11, %xmm14
+; SSE-NEXT:    movlhps {{.*#+}} xmm14 = xmm14[0],xmm10[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm12, %xmm11
+; SSE-NEXT:    movlhps {{.*#+}} xmm11 = xmm11[0],xmm15[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm15[1]
+; SSE-NEXT:    movaps %xmm12, %xmm15
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm10, %xmm0
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm9[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm10, %xmm9
+; SSE-NEXT:    movlhps {{.*#+}} xmm9 = xmm9[0],xmm8[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm8[1]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm10, %xmm8
+; SSE-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm7[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm7[1]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm7, %xmm12
+; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm6[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm6[1]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps (%rsp), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm5[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm5[1]
+; SSE-NEXT:    movaps %xmm6, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm3[1]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm10, %xmm3
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm13, %xmm6
+; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm1[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm1[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%r9)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rax)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm3, 240(%rax)
+; SSE-NEXT:    movaps %xmm4, 224(%rax)
+; SSE-NEXT:    movaps %xmm5, 208(%rax)
+; SSE-NEXT:    movaps %xmm7, 192(%rax)
+; SSE-NEXT:    movaps %xmm12, 176(%rax)
+; SSE-NEXT:    movaps %xmm8, 160(%rax)
+; SSE-NEXT:    movaps %xmm9, 144(%rax)
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps %xmm11, 112(%rax)
+; SSE-NEXT:    movaps %xmm14, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps %xmm6, (%rax)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm10, 240(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rax)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps %xmm15, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps %xmm13, (%rax)
+; SSE-NEXT:    addq $2232, %rsp # imm = 0x8B8
+; SSE-NEXT:    retq
+;
+; AVX1-ONLY-LABEL: load_i32_stride8_vf64:
+; AVX1-ONLY:       # %bb.0:
+; AVX1-ONLY-NEXT:    subq $3688, %rsp # imm = 0xE68
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm15
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm14[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 1504(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1472(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1312(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1280(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm13[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1952(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1920(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 2016(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1984(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1824(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1792(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps 1888(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1856(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm10[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vmovaps 1024(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm3[0],xmm12[0],xmm3[1],xmm12[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm9[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 1760(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1728(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm5[0],xmm8[0],xmm5[1],xmm8[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1632(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1600(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, %xmm11
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm7[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],xmm14[1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[0],mem[1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm15[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm13[0],mem[1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm10[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm13[1],xmm10[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm6[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm6 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm10[1],xmm6[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm4[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0],xmm6[1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm3[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm12[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vpermilps $85, (%rsp), %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[0],mem[1],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm8[1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm9[1,1,1,1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm5[1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm11[2],xmm7[2],xmm11[3],xmm7[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm0[2],xmm14[2],xmm0[3],xmm14[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm15[2],mem[2],xmm15[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm4[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm0[2],xmm13[2],xmm0[3],xmm13[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm0[2],xmm10[2],xmm0[3],xmm10[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm15[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm6[2],xmm0[3],xmm6[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = mem[0,1,2],xmm12[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm14[0,1,2,3,4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm13[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm1[0,1,2],xmm14[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm2[0,1],xmm14[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm14[0,1,2,3],ymm12[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm6 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm14[0,1,2],xmm12[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm10[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm7[0,1,2],xmm12[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm6[0,1],xmm12[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm12[0,1,2,3],ymm11[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm9[2],xmm5[2],xmm9[3],xmm5[3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm12[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm11[0,1,2],xmm9[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm5[0,1,2],xmm9[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm0[0,1],xmm9[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3],ymm8[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm2[1],xmm8[1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm9 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm15[2],mem[2],xmm15[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm1[2],xmm13[2],xmm1[3],xmm13[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm7[2],xmm10[2],xmm7[3],xmm10[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm14[2],mem[2],xmm14[3],mem[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm11[2],xmm12[2],xmm11[3],xmm12[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm6[0],ymm10[0],ymm6[2],ymm10[2]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm12[0],ymm2[2],ymm12[2]
+; AVX1-ONLY-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %ymm11
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm11[0],ymm1[0],ymm11[1],ymm1[1],ymm11[4],ymm1[4],ymm11[5],ymm1[5]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm3[0],ymm1[0],ymm3[1],ymm1[1],ymm3[4],ymm1[4],ymm3[5],ymm1[5]
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, %ymm5
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm9[0],ymm2[2],ymm9[2]
+; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1024(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1472(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1504(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 1312(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1280(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %ymm13
+; AVX1-ONLY-NEXT:    vmovaps 1728(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1760(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm13[0],ymm2[0],ymm13[1],ymm2[1],ymm13[4],ymm2[4],ymm13[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 1568(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1536(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm7
+; AVX1-ONLY-NEXT:    vmovaps 1600(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1632(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm14[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1952(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1920(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1984(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2016(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm7 = ymm1[0],ymm4[0],ymm1[1],ymm4[1],ymm1[4],ymm4[4],ymm1[5],ymm4[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm7[0,1],ymm0[2,0],ymm7[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 1824(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1792(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1856(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1888(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm14[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[0,1],ymm1[2,0],ymm8[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm14 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm10[0],ymm6[0],ymm10[1],ymm6[1],ymm10[4],ymm6[4],ymm10[5],ymm6[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm12[0],mem[0],ymm12[1],mem[1],ymm12[4],mem[4],ymm12[5],mem[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm1[1,0],ymm11[1,0],ymm1[5,4],ymm11[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm9[0],ymm3[0],ymm9[1],ymm3[1],ymm9[4],ymm3[4],ymm9[5],ymm3[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm1[1,0],ymm5[1,0],ymm1[5,4],ymm5[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[1],ymm4[1],ymm0[4],ymm4[4],ymm0[5],ymm4[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm11[0],ymm7[0],ymm11[1],ymm7[1],ymm11[4],ymm7[4],ymm11[5],ymm7[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm5[1,0],mem[1,0],ymm5[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm12[0],ymm5[0],ymm12[1],ymm5[1],ymm12[4],ymm5[4],ymm12[5],ymm5[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm1[1,0],ymm13[1,0],ymm1[5,4],ymm13[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm15[0],ymm13[0],ymm15[1],ymm13[1],ymm15[4],ymm13[4],ymm15[5],ymm13[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = ymm0[1,0],mem[1,0],ymm0[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm8[1,0],mem[1,0],ymm8[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm8[0],mem[0],ymm8[1],mem[1],ymm8[4],mem[4],ymm8[5],mem[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = ymm0[1,0],mem[1,0],ymm0[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm8[1,0],mem[1,0],ymm8[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm8[0],mem[0],ymm8[1],mem[1],ymm8[4],mem[4],ymm8[5],mem[5]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = ymm14[1,0],mem[1,0],ymm14[5,4],mem[5,4]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm6[1],ymm10[1],ymm6[3],ymm10[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm6[2],mem[2],ymm6[3],mem[3],ymm6[6],mem[6],ymm6[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,0],ymm8[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm10[1],ymm6[1],ymm10[3],ymm6[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,0],ymm8[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm3[1],ymm10[1],ymm3[3],ymm10[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm4[1],mem[1],ymm4[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm8 = ymm1[2],ymm4[2],ymm1[3],ymm4[3],ymm1[6],ymm4[6],ymm1[7],ymm4[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,0],ymm8[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm7[1],ymm11[1],ymm7[3],ymm11[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,0],ymm8[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm5[1],ymm12[1],ymm5[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,0],ymm8[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm9[1],ymm2[1],ymm9[3],ymm2[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,0],ymm8[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm13[1],ymm15[1],ymm13[3],ymm15[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm8 = ymm9[2],ymm7[2],ymm9[3],ymm7[3],ymm9[6],ymm7[6],ymm9[7],ymm7[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,0],ymm8[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps (%rsp), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm12[1],ymm0[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm8 = ymm11[2],ymm13[2],ymm11[3],ymm13[3],ymm11[6],ymm13[6],ymm11[7],ymm13[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,0],ymm8[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm3[1],mem[1],ymm3[3],mem[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = ymm15[2],mem[2],ymm15[3],mem[3],ymm15[6],mem[6],ymm15[7],mem[7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm6[2],mem[2],ymm6[3],mem[3],ymm6[6],mem[6],ymm6[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm8[0,1],xmm6[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm5 = ymm10[2],mem[2],ymm10[3],mem[3],ymm10[6],mem[6],ymm10[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm8[0,1],xmm5[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm4[3,0],mem[3,0],ymm4[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm8[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm2[2],mem[2],ymm2[3],mem[3],ymm2[6],mem[6],ymm2[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm8[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm8[3,0],mem[3,0],ymm8[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm7[3,0],ymm9[3,0],ymm7[7,4],ymm9[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm8[2,0],ymm0[2,3],ymm8[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = ymm7[2],mem[2],ymm7[3],mem[3],ymm7[6],mem[6],ymm7[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm8[3,0],mem[3,0],ymm8[7,4],mem[7,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm13[3,0],ymm11[3,0],ymm13[7,4],ymm11[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[2,0],ymm7[2,3],ymm8[6,4],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm8[2],mem[2],ymm8[3],mem[3],ymm8[6],mem[6],ymm8[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[3,0],ymm15[3,0],ymm9[7,4],ymm15[7,4]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm9, %xmm9
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%r9)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rax)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm14, (%rax)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm6, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, (%rax)
+; AVX1-ONLY-NEXT:    addq $3688, %rsp # imm = 0xE68
+; AVX1-ONLY-NEXT:    vzeroupper
+; AVX1-ONLY-NEXT:    retq
+;
+; AVX2-ONLY-LABEL: load_i32_stride8_vf64:
+; AVX2-ONLY:       # %bb.0:
+; AVX2-ONLY-NEXT:    subq $3272, %rsp # imm = 0xCC8
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm2, %xmm2
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm2, %xmm2
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm3, %xmm3
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 800(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 768(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps 864(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps 832(%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm11, %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 992(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps 960(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm2, %xmm2
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps 928(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 896(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm0, %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps 1312(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1280(%rdi), %xmm13
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 1504(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps 1472(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm2, %xmm2
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps 1440(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1888(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm0, %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1856(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps 1824(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1792(%rdi), %xmm15
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm1[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 2016(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm0, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps 1984(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm0, %xmm2
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps 1952(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1920(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 608(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm0, %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 576(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps 544(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 512(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 736(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps 704(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm2, %xmm2
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps 672(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 640(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1120(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm0, %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1088(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1024(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 1248(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm2, %xmm2
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1152(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1632(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm0, %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1600(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 1760(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps 1728(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm2, %xmm2
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps 1696(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1664(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm5, %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm7, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm6, %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vbroadcastss %xmm10, %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm14
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm14[0],xmm4[0],xmm14[1],xmm4[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm12[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm2[0],mem[1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm11[0],mem[0],xmm11[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm2[0],mem[1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm13[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm2[0],mem[1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm2[0],mem[1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm4[1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm10[0],xmm6[0],xmm10[1],xmm6[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm7[0],xmm5[0],xmm7[1],xmm5[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm8[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm9[1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm14[1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm8[0],xmm6[0],xmm8[1],xmm6[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm2[0],mem[1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm10[0],mem[0],xmm10[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm2[0],mem[1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm9[0],mem[0],xmm9[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm7[0],xmm5[0],xmm7[1],xmm5[1]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[1,1,1,1]
+; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm2[0],mem[1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps (%rsp), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm12[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[0,1,2],xmm1[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm11[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm11[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[0,1,2],xmm1[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm13[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[0,1,2],xmm1[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm15[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[0,1,2],xmm1[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm0[2],xmm14[2],xmm0[3],xmm14[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm6[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm8[0,1,2],xmm0[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm6[2,2,2,2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm8 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm8 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm4[2,2,2,2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm4[0,1,2],xmm0[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm10[0,1,2],xmm1[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm8 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm8 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm5[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm7[0,1,2],xmm0[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm15
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm1[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm9[0,1,2],xmm15[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm8[0,1],xmm15[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm5 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = mem[2,2,2,2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm0[0,1,2],xmm15[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm14
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm13 = mem[2,2,2,2]
+; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm13 = mem[0,1,2],xmm13[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm15[0,1],xmm13[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm13 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm13 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm13 = xmm5[1],xmm13[1]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm14 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm14 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-ONLY-NEXT:    vpermilps $238, (%rsp), %xmm12 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm12 = mem[2,3,2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm12 = xmm5[2],xmm11[2],xmm5[3],xmm11[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm11 = xmm5[1],xmm12[1]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm12 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm12 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm13 = mem[2,3,2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm11 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm11 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm9 = xmm5[1],xmm11[1]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm11 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm11 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
+; AVX2-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm10 = mem[2,3,2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm9 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm9 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm5[1],xmm9[1]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm9 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm9 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
+; AVX2-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm8 = mem[2,3,2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm7 = xmm3[2],xmm6[2],xmm3[3],xmm6[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm3[1],xmm7[1]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX2-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm6 = mem[2,3,2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm2[1],xmm5[1]
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm5 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm5 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm4 = mem[2,3,2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[2,3,2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = mem[2,3,2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm15[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 48(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpbroadcastd 16(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovdqa 96(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 64(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovdqa 224(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 192(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 160(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vmovdqa 128(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm11[0],ymm1[1],ymm11[1],ymm1[4],ymm11[4],ymm1[5],ymm11[5]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm11, %ymm10
+; AVX2-ONLY-NEXT:    vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm7 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
+; AVX2-ONLY-NEXT:    vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[2],ymm7[2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 304(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpbroadcastd 272(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovdqa 352(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 320(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vmovdqa 480(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 448(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 416(%rdi), %ymm12
+; AVX2-ONLY-NEXT:    vmovdqa 384(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm2 = ymm2[0],ymm12[0],ymm2[1],ymm12[1],ymm2[4],ymm12[4],ymm2[5],ymm12[5]
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm3[0],ymm1[0],ymm3[1],ymm1[1],ymm3[4],ymm1[4],ymm3[5],ymm1[5]
+; AVX2-ONLY-NEXT:    vpunpcklqdq {{.*#+}} ymm2 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 560(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpbroadcastd 528(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovdqa 608(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 576(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vmovdqa 736(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 704(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 672(%rdi), %ymm14
+; AVX2-ONLY-NEXT:    vmovdqa 640(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm3 = ymm3[0],ymm14[0],ymm3[1],ymm14[1],ymm3[4],ymm14[4],ymm3[5],ymm14[5]
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
+; AVX2-ONLY-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 816(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpbroadcastd 784(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovdqa 864(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 832(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT:    vmovdqa 992(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 960(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 928(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 896(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm4 = ymm4[0],ymm6[0],ymm4[1],ymm6[1],ymm4[4],ymm6[4],ymm4[5],ymm6[5]
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm3 = ymm5[0],ymm3[0],ymm5[1],ymm3[1],ymm5[4],ymm3[4],ymm5[5],ymm3[5]
+; AVX2-ONLY-NEXT:    vpunpcklqdq {{.*#+}} ymm4 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1072(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpbroadcastd 1040(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovdqa 1120(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1088(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm4 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[4],ymm5[4],ymm4[5],ymm5[5]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT:    vmovdqa 1248(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1216(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1184(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1152(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm5 = ymm5[0],ymm8[0],ymm5[1],ymm8[1],ymm5[4],ymm8[4],ymm5[5],ymm8[5]
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm4 = ymm6[0],ymm4[0],ymm6[1],ymm4[1],ymm6[4],ymm4[4],ymm6[5],ymm4[5]
+; AVX2-ONLY-NEXT:    vpunpcklqdq {{.*#+}} ymm5 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1328(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpbroadcastd 1296(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovdqa 1376(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1344(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm5 = ymm6[0],ymm5[0],ymm6[1],ymm5[1],ymm6[4],ymm5[4],ymm6[5],ymm5[5]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT:    vmovdqa 1504(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1472(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1440(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1408(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm6 = ymm6[0],ymm9[0],ymm6[1],ymm9[1],ymm6[4],ymm9[4],ymm6[5],ymm9[5]
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm5 = ymm8[0],ymm5[0],ymm8[1],ymm5[1],ymm8[4],ymm5[4],ymm8[5],ymm5[5]
+; AVX2-ONLY-NEXT:    vpunpcklqdq {{.*#+}} ymm6 = ymm6[0],ymm5[0],ymm6[2],ymm5[2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1584(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpbroadcastd 1552(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovdqa 1632(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1600(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm6 = ymm8[0],ymm6[0],ymm8[1],ymm6[1],ymm8[4],ymm6[4],ymm8[5],ymm6[5]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm6 = ymm6[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm6 = xmm0[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT:    vmovdqa 1760(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1728(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1696(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1664(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm8 = ymm8[0],ymm11[0],ymm8[1],ymm11[1],ymm8[4],ymm11[4],ymm8[5],ymm11[5]
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm13 = ymm9[0],ymm0[0],ymm9[1],ymm0[1],ymm9[4],ymm0[4],ymm9[5],ymm0[5]
+; AVX2-ONLY-NEXT:    vpunpcklqdq {{.*#+}} ymm8 = ymm8[0],ymm13[0],ymm8[2],ymm13[2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1840(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vpbroadcastd 1808(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1]
+; AVX2-ONLY-NEXT:    vmovdqa 1888(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1856(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovdqu %ymm9, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm8 = ymm9[0],ymm8[0],ymm9[1],ymm8[1],ymm9[4],ymm8[4],ymm9[5],ymm8[5]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm8 = xmm6[0,1],xmm8[2,3]
+; AVX2-ONLY-NEXT:    vmovdqa 2016(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1984(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1952(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vmovdqa 1920(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm9 = ymm0[0],ymm15[0],ymm0[1],ymm15[1],ymm0[4],ymm15[4],ymm0[5],ymm15[5]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} ymm6 = ymm11[0],ymm6[0],ymm11[1],ymm6[1],ymm11[4],ymm6[4],ymm11[5],ymm6[5]
+; AVX2-ONLY-NEXT:    vpunpcklqdq {{.*#+}} ymm9 = ymm9[0],ymm6[0],ymm9[2],ymm6[2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 148(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm7 = mem[0],zero,mem[1],zero
+; AVX2-ONLY-NEXT:    vpbroadcastd 116(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm9 = xmm7[0,1,2],xmm9[3]
+; AVX2-ONLY-NEXT:    vmovdqa 32(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm10 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0],ymm11[1],ymm10[2,3,4],ymm11[5],ymm10[6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm10, %xmm10
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 404(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm12[5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm8 = mem[0],zero,mem[1],zero
+; AVX2-ONLY-NEXT:    vpbroadcastd 372(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm8 = xmm8[0,1,2],xmm9[3]
+; AVX2-ONLY-NEXT:    vmovdqa 288(%rdi), %ymm10
+; AVX2-ONLY-NEXT:    vmovdqa 256(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm9 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2,3,4],ymm10[5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm9, %xmm9
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 660(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm14[5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm2 = mem[0],zero,mem[1],zero
+; AVX2-ONLY-NEXT:    vpbroadcastd 628(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm8[3]
+; AVX2-ONLY-NEXT:    vmovdqa 544(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 512(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm8 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0],ymm0[1],ymm8[2,3,4],ymm0[5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm8, %xmm8
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm8[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 916(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm8[5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm2 = mem[0],zero,mem[1],zero
+; AVX2-ONLY-NEXT:    vpbroadcastd 884(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3]
+; AVX2-ONLY-NEXT:    vmovdqa 800(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 768(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0],ymm0[1],ymm3[2,3,4],ymm0[5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm3, %xmm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1172(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm9[5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm2 = mem[0],zero,mem[1],zero
+; AVX2-ONLY-NEXT:    vpbroadcastd 1140(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3]
+; AVX2-ONLY-NEXT:    vmovdqa 1056(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1024(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0],ymm0[1],ymm3[2,3,4],ymm0[5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm3, %xmm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1428(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm7[5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm2 = mem[0],zero,mem[1],zero
+; AVX2-ONLY-NEXT:    vpbroadcastd 1396(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3]
+; AVX2-ONLY-NEXT:    vmovdqa 1312(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqa 1280(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2,3,4],ymm5[5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm3, %xmm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1684(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm4[5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero
+; AVX2-ONLY-NEXT:    vpbroadcastd 1652(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3]
+; AVX2-ONLY-NEXT:    vmovdqa 1568(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1536(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm2 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0],ymm13[1],ymm2[2,3,4],ymm13[5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1940(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm15[5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero
+; AVX2-ONLY-NEXT:    vpbroadcastd 1908(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3]
+; AVX2-ONLY-NEXT:    vmovdqa 1824(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vmovdqa 1792(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0],ymm15[1],ymm2[2,3,4],ymm15[5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 248(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm6 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm1[2],ymm11[2],ymm1[3],ymm11[3],ymm1[6],ymm11[6],ymm1[7],ymm11[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm2, %xmm1
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm3[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 504(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm3 = ymm1[2],ymm12[2],ymm1[3],ymm12[3],ymm1[6],ymm12[6],ymm1[7],ymm12[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm1[2],ymm10[2],ymm1[3],ymm10[3],ymm1[6],ymm10[6],ymm1[7],ymm10[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm6 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm2, %xmm1
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm6[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 760(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm3 = ymm1[2],ymm14[2],ymm1[3],ymm14[3],ymm1[6],ymm14[6],ymm1[7],ymm14[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm6 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm2, %xmm1
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm6[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1016(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm3 = ymm1[2],ymm8[2],ymm1[3],ymm8[3],ymm1[6],ymm8[6],ymm1[7],ymm8[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm12 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm2, %xmm1
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm12[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1272(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm13 = ymm1[2],ymm9[2],ymm1[3],ymm9[3],ymm1[6],ymm9[6],ymm1[7],ymm9[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm9 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm14, %xmm1
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm9[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1528(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm10 = ymm1[2],ymm7[2],ymm1[3],ymm7[3],ymm1[6],ymm7[6],ymm1[7],ymm7[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm11 = ymm1[2],ymm5[2],ymm1[3],ymm5[3],ymm1[6],ymm5[6],ymm1[7],ymm5[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm6 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm11, %xmm1
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm6[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 1784(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm7 = ymm1[2],ymm4[2],ymm1[3],ymm4[3],ymm1[6],ymm4[6],ymm1[7],ymm4[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm8, %xmm1
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm3[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastd 2040(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm0[2],ymm15[2],ymm0[3],ymm15[3],ymm0[6],ymm15[6],ymm0[7],ymm15[7]
+; AVX2-ONLY-NEXT:    vmovdqu (%rsp), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm5, %xmm0
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm15 = ymm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm15[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss 220(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss 476(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastss 732(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 988(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm12, %xmm1
+; AVX2-ONLY-NEXT:    vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm12 = mem[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm12, %xmm12
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm12[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm12 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 1244(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm13[1],ymm0[1],ymm13[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm9, %xmm1
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm9 = ymm14[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm9, %xmm9
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm9[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 1500(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm10[1],ymm0[1],ymm10[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm6, %xmm1
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm6 = ymm11[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm6, %xmm6
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 1756(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm7[1],ymm0[1],ymm7[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm3, %xmm1
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm8[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm3, %xmm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastd 2012(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm4[1],ymm0[1],ymm4[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm2, %xmm1
+; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm2 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vextracti128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%r9)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rax)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rax)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovdqa %ymm0, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm3, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm6, 160(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm9, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm12, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm15, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, (%rax)
+; AVX2-ONLY-NEXT:    addq $3272, %rsp # imm = 0xCC8
+; AVX2-ONLY-NEXT:    vzeroupper
+; AVX2-ONLY-NEXT:    retq
+;
+; AVX512F-LABEL: load_i32_stride8_vf64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    subq $3304, %rsp # imm = 0xCE8
+; AVX512F-NEXT:    vmovdqa64 1984(%rdi), %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1216(%rdi), %zmm22
+; AVX512F-NEXT:    vmovdqa64 1152(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 1088(%rdi), %zmm28
+; AVX512F-NEXT:    vmovdqa64 1024(%rdi), %zmm11
+; AVX512F-NEXT:    vmovdqa64 1344(%rdi), %zmm27
+; AVX512F-NEXT:    vmovdqa64 1280(%rdi), %zmm25
+; AVX512F-NEXT:    vmovdqa64 1472(%rdi), %zmm20
+; AVX512F-NEXT:    vmovdqa64 1408(%rdi), %zmm19
+; AVX512F-NEXT:    vmovdqa64 704(%rdi), %zmm8
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %zmm24
+; AVX512F-NEXT:    vmovdqa64 512(%rdi), %zmm10
+; AVX512F-NEXT:    vmovdqa64 832(%rdi), %zmm26
+; AVX512F-NEXT:    vmovdqa64 768(%rdi), %zmm31
+; AVX512F-NEXT:    vmovdqa64 960(%rdi), %zmm17
+; AVX512F-NEXT:    vmovdqa64 896(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqa64 (%rdi), %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 64(%rdi), %zmm15
+; AVX512F-NEXT:    vmovdqa64 128(%rdi), %zmm12
+; AVX512F-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 192(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqa64 320(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 256(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 448(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 384(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [0,8,16,24,0,8,16,24,0,8,16,24,0,8,16,24]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm4, %zmm0, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm3, %zmm0, %zmm2
+; AVX512F-NEXT:    movb $-64, %al
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,u,u,u,0,8,16,24>
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm4
+; AVX512F-NEXT:    vpermt2d %zmm5, %zmm3, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm30
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm21
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm4, %zmm2, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm23
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm17, %zmm15
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm4
+; AVX512F-NEXT:    vpermt2d %zmm26, %zmm0, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm4 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm24, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm24, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm8, %zmm3, %zmm5
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm19, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm25, %zmm4
+; AVX512F-NEXT:    vpermt2d %zmm27, %zmm0, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm4 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm28, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm22, %zmm3, %zmm5
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 1920(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm14, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 1856(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqa64 1792(%rdi), %zmm12
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm4, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm29
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqa64 1600(%rdi), %zmm16
+; AVX512F-NEXT:    vmovdqa64 1536(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm16, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm16, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1728(%rdi), %zmm9
+; AVX512F-NEXT:    vmovdqa64 1664(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2d %zmm9, %zmm1, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [1,9,17,25,1,9,17,25,1,9,17,25,1,9,17,25]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm19, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm25, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm27, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm27, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512F-NEXT:    vpermt2d %zmm28, %zmm0, %zmm11
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm28, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,1,9,17,25>
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm22, %zmm1, %zmm8
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm5 = ymm11[0,1,2,3],ymm8[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm5, %zmm2, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm31, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm26, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm24, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm7, %zmm1, %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm18
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm5
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm22, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm30, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm30, %zmm1, %zmm2
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm21, %zmm0, %zmm8
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm7, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2d %zmm16, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2d %zmm9, %zmm2, %zmm1
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [2,10,18,26,2,10,18,26,2,10,18,26,2,10,18,26]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm0, %zmm19
+; AVX512F-NEXT:    vmovdqa64 %zmm25, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm27, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm19, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm14, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,2,10,18,26>
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm3, %zmm1, %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm27
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm5, %zmm2, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm23, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm26, %zmm0, %zmm31
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm31 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm16, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm14, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm1, %zmm8
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm31, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm17, %zmm23
+; AVX512F-NEXT:    vpermt2d %zmm17, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm17
+; AVX512F-NEXT:    vpermt2d %zmm22, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512F-NEXT:    vpermt2d %zmm30, %zmm1, %zmm4
+; AVX512F-NEXT:    vpermt2d %zmm21, %zmm0, %zmm13
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm7, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm19
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm6
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm29, %zmm31
+; AVX512F-NEXT:    vmovdqu64 %zmm29, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2d %zmm13, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2d %zmm4, %zmm2, %zmm1
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [3,11,19,27,3,11,19,27,3,11,19,27,3,11,19,27]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm25, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm5 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,3,11,19,27>
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm27, %zmm1, %zmm8
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm5, %zmm2, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm15, %zmm0, %zmm9
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm26, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm26, %zmm24
+; AVX512F-NEXT:    vmovdqu64 %zmm26, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm5 {%k1}
+; AVX512F-NEXT:    vpermt2d %zmm16, %zmm0, %zmm10
+; AVX512F-NEXT:    vmovdqa64 %zmm14, %zmm8
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm30, %zmm1, %zmm8
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm10[0,1,2,3],ymm8[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm11
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm23, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm28
+; AVX512F-NEXT:    vmovdqa64 %zmm17, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm22, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm23
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm14, %zmm1, %zmm2
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm21, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm27
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm26
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm19, %zmm29
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm31, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2d %zmm13, %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm21
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2d %zmm4, %zmm15, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm6
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [4,12,20,28,4,12,20,28,4,12,20,28,4,12,20,28]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm1
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm20, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm25, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm25, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm5
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm16, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,4,12,20,28>
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm8
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm1, %zmm8
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm5, %zmm2, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512F-NEXT:    vpermt2d {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm24, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm24, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm30, %zmm1, %zmm8
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm28, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm17, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm23, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm14, %zmm1, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm27, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm9
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm26, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm26, %zmm23
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm27, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm22 {%k1}
+; AVX512F-NEXT:    vpermi2d %zmm21, %zmm3, %zmm0
+; AVX512F-NEXT:    vpermi2d %zmm6, %zmm15, %zmm1
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm22, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm3 = [5,13,21,29,5,13,21,29,5,13,21,29,5,13,21,29]
+; AVX512F-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm3, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm18, %zmm3, %zmm25
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm25 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm2
+; AVX512F-NEXT:    vpermt2d %zmm16, %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,5,13,21,29>
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm19, %zmm1, %zmm8
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm8[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm25, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm24, %zmm3, %zmm0
+; AVX512F-NEXT:    vpermt2d %zmm30, %zmm1, %zmm7
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm16
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm3, %zmm16
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm5 = [6,14,22,30,6,14,22,30,6,14,22,30,6,14,22,30]
+; AVX512F-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm4
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm5, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm25 = [7,15,23,31,7,15,23,31,7,15,23,31,7,15,23,31]
+; AVX512F-NEXT:    # zmm25 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm25, %zmm31
+; AVX512F-NEXT:    vmovdqu64 %zmm31, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm7
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm3, %zmm7
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm19
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm5, %zmm19
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm25, %zmm30
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm4
+; AVX512F-NEXT:    vpermt2d %zmm24, %zmm5, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm24, %zmm25, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm18
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm26
+; AVX512F-NEXT:    vpermt2d %zmm28, %zmm3, %zmm26
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm17
+; AVX512F-NEXT:    vpermt2d %zmm28, %zmm5, %zmm17
+; AVX512F-NEXT:    vpermt2d %zmm28, %zmm25, %zmm18
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm28
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm3, %zmm28
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm11
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm5, %zmm11
+; AVX512F-NEXT:    vpermt2d %zmm0, %zmm25, %zmm21
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm15
+; AVX512F-NEXT:    vpermt2d %zmm9, %zmm3, %zmm15
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm9, %zmm5, %zmm8
+; AVX512F-NEXT:    vpermt2d %zmm9, %zmm25, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm22
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm3, %zmm22
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm5, %zmm23
+; AVX512F-NEXT:    vpermt2d %zmm29, %zmm25, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm29
+; AVX512F-NEXT:    vpermt2d %zmm27, %zmm3, %zmm29
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm24
+; AVX512F-NEXT:    vpermt2d %zmm27, %zmm5, %zmm24
+; AVX512F-NEXT:    vpermt2d %zmm27, %zmm25, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2d %zmm12, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm9
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm31
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm27, %zmm5, %zmm31
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm10
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm5, %zmm10
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm4
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm5, %zmm4
+; AVX512F-NEXT:    vpermi2d %zmm12, %zmm0, %zmm5
+; AVX512F-NEXT:    vpermt2d %zmm12, %zmm25, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2d %zmm27, %zmm25, %zmm9
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm27
+; AVX512F-NEXT:    vpermt2d %zmm20, %zmm25, %zmm6
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm20
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm25, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm7 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm25 # 32-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm26, %zmm28 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm14, %zmm12
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,5,13,21,29>
+; AVX512F-NEXT:    vpermt2d %zmm26, %zmm1, %zmm12
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm12 = ymm15[0,1,2,3],ymm12[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm12, %zmm28, %zmm28
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm29 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2d %zmm22, %zmm0, %zmm1
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm29, %zmm29
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,u,u,u,6,14,22,30>
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm15
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm3, %zmm15
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} ymm16 = <u,u,u,u,7,15,23,31>
+; AVX512F-NEXT:    vpermt2d %zmm2, %zmm16, %zmm9
+; AVX512F-NEXT:    vmovdqa64 %zmm14, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm26, %zmm3, %zmm7
+; AVX512F-NEXT:    vpermt2d %zmm26, %zmm16, %zmm14
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm3, %zmm2
+; AVX512F-NEXT:    vpermi2d %zmm22, %zmm0, %zmm3
+; AVX512F-NEXT:    vpermt2d %zmm22, %zmm16, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm6
+; AVX512F-NEXT:    vpermt2d %zmm1, %zmm16, %zmm12
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm10 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm10, %zmm1
+; AVX512F-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm10 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm10 = mem[0,1,2,3],ymm15[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm19 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm10, %zmm19, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm17, %zmm11 {%k1}
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm7, %zmm11, %zmm7
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm24 {%k1}
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm24, %zmm0
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm30 {%k1}
+; AVX512F-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm3 = mem[0,1,2,3],ymm9[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm3, %zmm30, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm21 {%k1}
+; AVX512F-NEXT:    vpblendd $15, (%rsp), %ymm14, %ymm8 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm8 = mem[0,1,2,3],ymm14[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm8, %zmm21, %zmm9
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm13 {%k1}
+; AVX512F-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm8 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm8 = mem[0,1,2,3],ymm6[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm8, %zmm13, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm20 {%k1}
+; AVX512F-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm8 = mem[0,1,2,3],ymm12[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm8, %zmm20, %zmm8
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 192(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 128(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 64(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, (%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 192(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, (%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 64(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 128(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 192(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, (%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 64(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 128(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 192(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, (%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 64(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 128(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 192(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, (%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 64(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 128(%r9)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovdqa64 %zmm29, 192(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm28, (%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm25, 64(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm5, 128(%rax)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovdqa64 %zmm0, 192(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm7, (%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm2, 64(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm1, 128(%rax)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovdqa64 %zmm8, 128(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm4, 192(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm9, (%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm3, 64(%rax)
+; AVX512F-NEXT:    addq $3304, %rsp # imm = 0xCE8
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: load_i32_stride8_vf64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    subq $3304, %rsp # imm = 0xCE8
+; AVX512BW-NEXT:    vmovdqa64 1984(%rdi), %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %zmm22
+; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %zmm28
+; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %zmm11
+; AVX512BW-NEXT:    vmovdqa64 1344(%rdi), %zmm27
+; AVX512BW-NEXT:    vmovdqa64 1280(%rdi), %zmm25
+; AVX512BW-NEXT:    vmovdqa64 1472(%rdi), %zmm20
+; AVX512BW-NEXT:    vmovdqa64 1408(%rdi), %zmm19
+; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm8
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm24
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm10
+; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm26
+; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm31
+; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm17
+; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm15
+; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm12
+; AVX512BW-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm5
+; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [0,8,16,24,0,8,16,24,0,8,16,24,0,8,16,24]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm4, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm0, %zmm2
+; AVX512BW-NEXT:    movb $-64, %al
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,u,u,u,0,8,16,24>
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm4
+; AVX512BW-NEXT:    vpermt2d %zmm5, %zmm3, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm30
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm21
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm4, %zmm2, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm23
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, %zmm15
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm4
+; AVX512BW-NEXT:    vpermt2d %zmm26, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm4 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm24, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm24, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm8, %zmm3, %zmm5
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm4
+; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm4 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm28, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm22, %zmm3, %zmm5
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 1920(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm14, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 1856(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqa64 1792(%rdi), %zmm12
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm4, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm29
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 1600(%rdi), %zmm16
+; AVX512BW-NEXT:    vmovdqa64 1536(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm16, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm16, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1728(%rdi), %zmm9
+; AVX512BW-NEXT:    vmovdqa64 1664(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2d %zmm9, %zmm1, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [1,9,17,25,1,9,17,25,1,9,17,25,1,9,17,25]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm27, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512BW-NEXT:    vpermt2d %zmm28, %zmm0, %zmm11
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm28, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,1,9,17,25>
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm22, %zmm1, %zmm8
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm11[0,1,2,3],ymm8[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm5, %zmm2, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm31, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm26, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm24, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm7, %zmm1, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm18
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm22, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm30, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm30, %zmm1, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm21, %zmm0, %zmm8
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3],ymm2[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm7, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2d %zmm16, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2d %zmm9, %zmm2, %zmm1
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [2,10,18,26,2,10,18,26,2,10,18,26,2,10,18,26]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm0, %zmm19
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm14, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,2,10,18,26>
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm1, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm27
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm5, %zmm2, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm23, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm26, %zmm0, %zmm31
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm31 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm16, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm1, %zmm8
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm31, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, %zmm23
+; AVX512BW-NEXT:    vpermt2d %zmm17, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm17
+; AVX512BW-NEXT:    vpermt2d %zmm22, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512BW-NEXT:    vpermt2d %zmm30, %zmm1, %zmm4
+; AVX512BW-NEXT:    vpermt2d %zmm21, %zmm0, %zmm13
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm7, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm19
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm6
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, %zmm31
+; AVX512BW-NEXT:    vmovdqu64 %zmm29, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2d %zmm13, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2d %zmm4, %zmm2, %zmm1
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [3,11,19,27,3,11,19,27,3,11,19,27,3,11,19,27]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm5 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,3,11,19,27>
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm1, %zmm8
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm5, %zmm2, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm15, %zmm0, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm26, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm24
+; AVX512BW-NEXT:    vmovdqu64 %zmm26, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm5 {%k1}
+; AVX512BW-NEXT:    vpermt2d %zmm16, %zmm0, %zmm10
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm30, %zmm1, %zmm8
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm10[0,1,2,3],ymm8[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm11
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm23, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm28
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm22, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm23
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm14, %zmm1, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm21, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm27
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3],ymm2[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm26
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm29
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm22
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm31, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2d %zmm13, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm21
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2d %zmm4, %zmm15, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm6
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [4,12,20,28,4,12,20,28,4,12,20,28,4,12,20,28]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm1
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm20, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm25, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm16, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,4,12,20,28>
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm1, %zmm8
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm5, %zmm2, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512BW-NEXT:    vpermt2d {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm24, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm24, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm30, %zmm1, %zmm8
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm28, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm23, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm14, %zmm1, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm9
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3],ymm2[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm23
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm0, %zmm22
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm22 {%k1}
+; AVX512BW-NEXT:    vpermi2d %zmm21, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpermi2d %zmm6, %zmm15, %zmm1
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm22, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm3 = [5,13,21,29,5,13,21,29,5,13,21,29,5,13,21,29]
+; AVX512BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm3, %zmm25
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm25 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm2
+; AVX512BW-NEXT:    vpermt2d %zmm16, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,5,13,21,29>
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm19, %zmm1, %zmm8
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm8[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm25, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm24, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpermt2d %zmm30, %zmm1, %zmm7
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm16
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm3, %zmm16
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm5 = [6,14,22,30,6,14,22,30,6,14,22,30,6,14,22,30]
+; AVX512BW-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm4
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm5, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm25 = [7,15,23,31,7,15,23,31,7,15,23,31,7,15,23,31]
+; AVX512BW-NEXT:    # zmm25 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm25, %zmm31
+; AVX512BW-NEXT:    vmovdqu64 %zmm31, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm3, %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm19
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm5, %zmm19
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm25, %zmm30
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm4
+; AVX512BW-NEXT:    vpermt2d %zmm24, %zmm5, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm24, %zmm25, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm18
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm26
+; AVX512BW-NEXT:    vpermt2d %zmm28, %zmm3, %zmm26
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm17
+; AVX512BW-NEXT:    vpermt2d %zmm28, %zmm5, %zmm17
+; AVX512BW-NEXT:    vpermt2d %zmm28, %zmm25, %zmm18
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm28
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm3, %zmm28
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm11
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm5, %zmm11
+; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm25, %zmm21
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm15
+; AVX512BW-NEXT:    vpermt2d %zmm9, %zmm3, %zmm15
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm9, %zmm5, %zmm8
+; AVX512BW-NEXT:    vpermt2d %zmm9, %zmm25, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm22
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm3, %zmm22
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm5, %zmm23
+; AVX512BW-NEXT:    vpermt2d %zmm29, %zmm25, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm29
+; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm3, %zmm29
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm24
+; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm5, %zmm24
+; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm25, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2d %zmm12, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm9
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm31
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm5, %zmm31
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm10
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm5, %zmm10
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm5, %zmm4
+; AVX512BW-NEXT:    vpermi2d %zmm12, %zmm0, %zmm5
+; AVX512BW-NEXT:    vpermt2d %zmm12, %zmm25, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm25, %zmm9
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm27
+; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm25, %zmm6
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm20
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm25, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm7 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm25 # 32-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm28 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, %zmm12
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,u,u,5,13,21,29>
+; AVX512BW-NEXT:    vpermt2d %zmm26, %zmm1, %zmm12
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm12 = ymm15[0,1,2,3],ymm12[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm12, %zmm28, %zmm28
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm29 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2d %zmm22, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm29, %zmm29
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,u,u,u,6,14,22,30>
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm15
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm3, %zmm15
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} ymm16 = <u,u,u,u,7,15,23,31>
+; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm16, %zmm9
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm26, %zmm3, %zmm7
+; AVX512BW-NEXT:    vpermt2d %zmm26, %zmm16, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpermi2d %zmm22, %zmm0, %zmm3
+; AVX512BW-NEXT:    vpermt2d %zmm22, %zmm16, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm6
+; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm16, %zmm12
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm10 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm10, %zmm1
+; AVX512BW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm10 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm10 = mem[0,1,2,3],ymm15[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm19 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm10, %zmm19, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, %zmm11 {%k1}
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm7, %zmm11, %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm24 {%k1}
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm24, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm30 {%k1}
+; AVX512BW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm3 = mem[0,1,2,3],ymm9[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm3, %zmm30, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm21 {%k1}
+; AVX512BW-NEXT:    vpblendd $15, (%rsp), %ymm14, %ymm8 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm8 = mem[0,1,2,3],ymm14[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm8, %zmm21, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm13 {%k1}
+; AVX512BW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm8 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm8 = mem[0,1,2,3],ymm6[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm8, %zmm13, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm20 {%k1}
+; AVX512BW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm8 = mem[0,1,2,3],ymm12[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm8, %zmm20, %zmm8
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 192(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 128(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 64(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, (%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 192(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, (%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 64(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 128(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 192(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, (%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 64(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 128(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 192(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, (%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 64(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 128(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 192(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, (%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 64(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 128(%r9)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, 192(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, (%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, 64(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm5, 128(%rax)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, 192(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, (%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, 64(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, 128(%rax)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, 128(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, 192(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, (%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, 64(%rax)
+; AVX512BW-NEXT:    addq $3304, %rsp # imm = 0xCE8
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %wide.vec = load <512 x i32>, ptr %in.vec, align 64
+  %strided.vec0 = shufflevector <512 x i32> %wide.vec, <512 x i32> poison, <64 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120, i32 128, i32 136, i32 144, i32 152, i32 160, i32 168, i32 176, i32 184, i32 192, i32 200, i32 208, i32 216, i32 224, i32 232, i32 240, i32 248, i32 256, i32 264, i32 272, i32 280, i32 288, i32 296, i32 304, i32 312, i32 320, i32 328, i32 336, i32 344, i32 352, i32 360, i32 368, i32 376, i32 384, i32 392, i32 400, i32 408, i32 416, i32 424, i32 432, i32 440, i32 448, i32 456, i32 464, i32 472, i32 480, i32 488, i32 496, i32 504>
+  %strided.vec1 = shufflevector <512 x i32> %wide.vec, <512 x i32> poison, <64 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121, i32 129, i32 137, i32 145, i32 153, i32 161, i32 169, i32 177, i32 185, i32 193, i32 201, i32 209, i32 217, i32 225, i32 233, i32 241, i32 249, i32 257, i32 265, i32 273, i32 281, i32 289, i32 297, i32 305, i32 313, i32 321, i32 329, i32 337, i32 345, i32 353, i32 361, i32 369, i32 377, i32 385, i32 393, i32 401, i32 409, i32 417, i32 425, i32 433, i32 441, i32 449, i32 457, i32 465, i32 473, i32 481, i32 489, i32 497, i32 505>
+  %strided.vec2 = shufflevector <512 x i32> %wide.vec, <512 x i32> poison, <64 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122, i32 130, i32 138, i32 146, i32 154, i32 162, i32 170, i32 178, i32 186, i32 194, i32 202, i32 210, i32 218, i32 226, i32 234, i32 242, i32 250, i32 258, i32 266, i32 274, i32 282, i32 290, i32 298, i32 306, i32 314, i32 322, i32 330, i32 338, i32 346, i32 354, i32 362, i32 370, i32 378, i32 386, i32 394, i32 402, i32 410, i32 418, i32 426, i32 434, i32 442, i32 450, i32 458, i32 466, i32 474, i32 482, i32 490, i32 498, i32 506>
+  %strided.vec3 = shufflevector <512 x i32> %wide.vec, <512 x i32> poison, <64 x i32> <i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 51, i32 59, i32 67, i32 75, i32 83, i32 91, i32 99, i32 107, i32 115, i32 123, i32 131, i32 139, i32 147, i32 155, i32 163, i32 171, i32 179, i32 187, i32 195, i32 203, i32 211, i32 219, i32 227, i32 235, i32 243, i32 251, i32 259, i32 267, i32 275, i32 283, i32 291, i32 299, i32 307, i32 315, i32 323, i32 331, i32 339, i32 347, i32 355, i32 363, i32 371, i32 379, i32 387, i32 395, i32 403, i32 411, i32 419, i32 427, i32 435, i32 443, i32 451, i32 459, i32 467, i32 475, i32 483, i32 491, i32 499, i32 507>
+  %strided.vec4 = shufflevector <512 x i32> %wide.vec, <512 x i32> poison, <64 x i32> <i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 52, i32 60, i32 68, i32 76, i32 84, i32 92, i32 100, i32 108, i32 116, i32 124, i32 132, i32 140, i32 148, i32 156, i32 164, i32 172, i32 180, i32 188, i32 196, i32 204, i32 212, i32 220, i32 228, i32 236, i32 244, i32 252, i32 260, i32 268, i32 276, i32 284, i32 292, i32 300, i32 308, i32 316, i32 324, i32 332, i32 340, i32 348, i32 356, i32 364, i32 372, i32 380, i32 388, i32 396, i32 404, i32 412, i32 420, i32 428, i32 436, i32 444, i32 452, i32 460, i32 468, i32 476, i32 484, i32 492, i32 500, i32 508>
+  %strided.vec5 = shufflevector <512 x i32> %wide.vec, <512 x i32> poison, <64 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125, i32 133, i32 141, i32 149, i32 157, i32 165, i32 173, i32 181, i32 189, i32 197, i32 205, i32 213, i32 221, i32 229, i32 237, i32 245, i32 253, i32 261, i32 269, i32 277, i32 285, i32 293, i32 301, i32 309, i32 317, i32 325, i32 333, i32 341, i32 349, i32 357, i32 365, i32 373, i32 381, i32 389, i32 397, i32 405, i32 413, i32 421, i32 429, i32 437, i32 445, i32 453, i32 461, i32 469, i32 477, i32 485, i32 493, i32 501, i32 509>
+  %strided.vec6 = shufflevector <512 x i32> %wide.vec, <512 x i32> poison, <64 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126, i32 134, i32 142, i32 150, i32 158, i32 166, i32 174, i32 182, i32 190, i32 198, i32 206, i32 214, i32 222, i32 230, i32 238, i32 246, i32 254, i32 262, i32 270, i32 278, i32 286, i32 294, i32 302, i32 310, i32 318, i32 326, i32 334, i32 342, i32 350, i32 358, i32 366, i32 374, i32 382, i32 390, i32 398, i32 406, i32 414, i32 422, i32 430, i32 438, i32 446, i32 454, i32 462, i32 470, i32 478, i32 486, i32 494, i32 502, i32 510>
+  %strided.vec7 = shufflevector <512 x i32> %wide.vec, <512 x i32> poison, <64 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127, i32 135, i32 143, i32 151, i32 159, i32 167, i32 175, i32 183, i32 191, i32 199, i32 207, i32 215, i32 223, i32 231, i32 239, i32 247, i32 255, i32 263, i32 271, i32 279, i32 287, i32 295, i32 303, i32 311, i32 319, i32 327, i32 335, i32 343, i32 351, i32 359, i32 367, i32 375, i32 383, i32 391, i32 399, i32 407, i32 415, i32 423, i32 431, i32 439, i32 447, i32 455, i32 463, i32 471, i32 479, i32 487, i32 495, i32 503, i32 511>
+  store <64 x i32> %strided.vec0, ptr %out.vec0, align 64
+  store <64 x i32> %strided.vec1, ptr %out.vec1, align 64
+  store <64 x i32> %strided.vec2, ptr %out.vec2, align 64
+  store <64 x i32> %strided.vec3, ptr %out.vec3, align 64
+  store <64 x i32> %strided.vec4, ptr %out.vec4, align 64
+  store <64 x i32> %strided.vec5, ptr %out.vec5, align 64
+  store <64 x i32> %strided.vec6, ptr %out.vec6, align 64
+  store <64 x i32> %strided.vec7, ptr %out.vec7, align 64
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll
index af4ae23..9df756a 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll
@@ -37,13 +37,10 @@ define void @load_i64_stride2_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nou
 ; AVX-NEXT:    vmovaps %xmm0, (%rdx)
 ; AVX-NEXT:    retq
   %wide.vec = load <4 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <4 x i64> %wide.vec, <4 x i64> poison, <2 x i32> <i32 0, i32 2>
   %strided.vec1 = shufflevector <4 x i64> %wide.vec, <4 x i64> poison, <2 x i32> <i32 1, i32 3>
-
   store <2 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i64> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -143,13 +140,10 @@ define void @load_i64_stride2_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nou
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <8 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <8 x i64> %wide.vec, <8 x i64> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
   %strided.vec1 = shufflevector <8 x i64> %wide.vec, <8 x i64> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
-
   store <4 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i64> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -239,13 +233,10 @@ define void @load_i64_stride2_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nou
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <16 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %strided.vec1 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
-
   store <8 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i64> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -402,13 +393,10 @@ define void @load_i64_stride2_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) no
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <32 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %strided.vec1 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
-
   store <16 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i64> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -733,13 +721,10 @@ define void @load_i64_stride2_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) no
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <64 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46, i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62>
   %strided.vec1 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <32 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31, i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47, i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63>
-
   store <32 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i64> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -1456,13 +1441,10 @@ define void @load_i64_stride2_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) no
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <128 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <64 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46, i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62, i32 64, i32 66, i32 68, i32 70, i32 72, i32 74, i32 76, i32 78, i32 80, i32 82, i32 84, i32 86, i32 88, i32 90, i32 92, i32 94, i32 96, i32 98, i32 100, i32 102, i32 104, i32 106, i32 108, i32 110, i32 112, i32 114, i32 116, i32 118, i32 120, i32 122, i32 124, i32 126>
   %strided.vec1 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <64 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31, i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47, i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63, i32 65, i32 67, i32 69, i32 71, i32 73, i32 75, i32 77, i32 79, i32 81, i32 83, i32 85, i32 87, i32 89, i32 91, i32 93, i32 95, i32 97, i32 99, i32 101, i32 103, i32 105, i32 107, i32 109, i32 111, i32 113, i32 115, i32 117, i32 119, i32 121, i32 123, i32 125, i32 127>
-
   store <64 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i64> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
index ab72571..76d4834 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
@@ -108,15 +108,12 @@ define void @load_i64_stride3_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <6 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <6 x i64> %wide.vec, <6 x i64> poison, <2 x i32> <i32 0, i32 3>
   %strided.vec1 = shufflevector <6 x i64> %wide.vec, <6 x i64> poison, <2 x i32> <i32 1, i32 4>
   %strided.vec2 = shufflevector <6 x i64> %wide.vec, <6 x i64> poison, <2 x i32> <i32 2, i32 5>
-
   store <2 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i64> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -203,15 +200,12 @@ define void @load_i64_stride3_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <12 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <12 x i64> %wide.vec, <12 x i64> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
   %strided.vec1 = shufflevector <12 x i64> %wide.vec, <12 x i64> poison, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
   %strided.vec2 = shufflevector <12 x i64> %wide.vec, <12 x i64> poison, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
-
   store <4 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i64> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -357,15 +351,12 @@ define void @load_i64_stride3_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <24 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <24 x i64> %wide.vec, <24 x i64> poison, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
   %strided.vec1 = shufflevector <24 x i64> %wide.vec, <24 x i64> poison, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
   %strided.vec2 = shufflevector <24 x i64> %wide.vec, <24 x i64> poison, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
-
   store <8 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i64> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -649,15 +640,12 @@ define void @load_i64_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <48 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <48 x i64> %wide.vec, <48 x i64> poison, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
   %strided.vec1 = shufflevector <48 x i64> %wide.vec, <48 x i64> poison, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
   %strided.vec2 = shufflevector <48 x i64> %wide.vec, <48 x i64> poison, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
-
   store <16 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i64> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -1281,15 +1269,12 @@ define void @load_i64_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <96 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <96 x i64> %wide.vec, <96 x i64> poison, <32 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45, i32 48, i32 51, i32 54, i32 57, i32 60, i32 63, i32 66, i32 69, i32 72, i32 75, i32 78, i32 81, i32 84, i32 87, i32 90, i32 93>
   %strided.vec1 = shufflevector <96 x i64> %wide.vec, <96 x i64> poison, <32 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46, i32 49, i32 52, i32 55, i32 58, i32 61, i32 64, i32 67, i32 70, i32 73, i32 76, i32 79, i32 82, i32 85, i32 88, i32 91, i32 94>
   %strided.vec2 = shufflevector <96 x i64> %wide.vec, <96 x i64> poison, <32 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47, i32 50, i32 53, i32 56, i32 59, i32 62, i32 65, i32 68, i32 71, i32 74, i32 77, i32 80, i32 83, i32 86, i32 89, i32 92, i32 95>
-
   store <32 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i64> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -2652,15 +2637,12 @@ define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <192 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <192 x i64> %wide.vec, <192 x i64> poison, <64 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45, i32 48, i32 51, i32 54, i32 57, i32 60, i32 63, i32 66, i32 69, i32 72, i32 75, i32 78, i32 81, i32 84, i32 87, i32 90, i32 93, i32 96, i32 99, i32 102, i32 105, i32 108, i32 111, i32 114, i32 117, i32 120, i32 123, i32 126, i32 129, i32 132, i32 135, i32 138, i32 141, i32 144, i32 147, i32 150, i32 153, i32 156, i32 159, i32 162, i32 165, i32 168, i32 171, i32 174, i32 177, i32 180, i32 183, i32 186, i32 189>
   %strided.vec1 = shufflevector <192 x i64> %wide.vec, <192 x i64> poison, <64 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46, i32 49, i32 52, i32 55, i32 58, i32 61, i32 64, i32 67, i32 70, i32 73, i32 76, i32 79, i32 82, i32 85, i32 88, i32 91, i32 94, i32 97, i32 100, i32 103, i32 106, i32 109, i32 112, i32 115, i32 118, i32 121, i32 124, i32 127, i32 130, i32 133, i32 136, i32 139, i32 142, i32 145, i32 148, i32 151, i32 154, i32 157, i32 160, i32 163, i32 166, i32 169, i32 172, i32 175, i32 178, i32 181, i32 184, i32 187, i32 190>
   %strided.vec2 = shufflevector <192 x i64> %wide.vec, <192 x i64> poison, <64 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47, i32 50, i32 53, i32 56, i32 59, i32 62, i32 65, i32 68, i32 71, i32 74, i32 77, i32 80, i32 83, i32 86, i32 89, i32 92, i32 95, i32 98, i32 101, i32 104, i32 107, i32 110, i32 113, i32 116, i32 119, i32 122, i32 125, i32 128, i32 131, i32 134, i32 137, i32 140, i32 143, i32 146, i32 149, i32 152, i32 155, i32 158, i32 161, i32 164, i32 167, i32 170, i32 173, i32 176, i32 179, i32 182, i32 185, i32 188, i32 191>
-
   store <64 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i64> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll
index 2c2a029..480e382 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll
@@ -135,17 +135,14 @@ define void @load_i64_stride4_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <8 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <8 x i64> %wide.vec, <8 x i64> poison, <2 x i32> <i32 0, i32 4>
   %strided.vec1 = shufflevector <8 x i64> %wide.vec, <8 x i64> poison, <2 x i32> <i32 1, i32 5>
   %strided.vec2 = shufflevector <8 x i64> %wide.vec, <8 x i64> poison, <2 x i32> <i32 2, i32 6>
   %strided.vec3 = shufflevector <8 x i64> %wide.vec, <8 x i64> poison, <2 x i32> <i32 3, i32 7>
-
   store <2 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <2 x i64> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -243,17 +240,14 @@ define void @load_i64_stride4_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <16 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
   %strided.vec1 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
   %strided.vec2 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
   %strided.vec3 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
-
   store <4 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <4 x i64> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -474,17 +468,14 @@ define void @load_i64_stride4_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <32 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
   %strided.vec1 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
   %strided.vec2 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
   %strided.vec3 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
-
   store <8 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <8 x i64> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -995,17 +986,14 @@ define void @load_i64_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <64 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
   %strided.vec1 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
   %strided.vec2 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
   %strided.vec3 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
-
   store <16 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <16 x i64> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -2086,17 +2074,2310 @@ define void @load_i64_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <128 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <32 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124>
   %strided.vec1 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <32 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125>
   %strided.vec2 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <32 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126>
   %strided.vec3 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <32 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127>
-
   store <32 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <32 x i64> %strided.vec3, ptr %out.vec3, align 64
+  ret void
+}
 
+define void @load_i64_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
+; SSE-LABEL: load_i64_stride4_vf64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    subq $1688, %rsp # imm = 0x698
+; SSE-NEXT:    movaps 544(%rdi), %xmm0
+; SSE-NEXT:    movaps 480(%rdi), %xmm1
+; SSE-NEXT:    movaps 448(%rdi), %xmm8
+; SSE-NEXT:    movaps 416(%rdi), %xmm2
+; SSE-NEXT:    movaps 384(%rdi), %xmm9
+; SSE-NEXT:    movaps 352(%rdi), %xmm3
+; SSE-NEXT:    movaps 320(%rdi), %xmm10
+; SSE-NEXT:    movaps 288(%rdi), %xmm4
+; SSE-NEXT:    movaps 256(%rdi), %xmm11
+; SSE-NEXT:    movaps 224(%rdi), %xmm5
+; SSE-NEXT:    movaps 192(%rdi), %xmm12
+; SSE-NEXT:    movaps 160(%rdi), %xmm6
+; SSE-NEXT:    movaps 128(%rdi), %xmm13
+; SSE-NEXT:    movaps 96(%rdi), %xmm7
+; SSE-NEXT:    movaps 64(%rdi), %xmm14
+; SSE-NEXT:    movaps %xmm14, %xmm15
+; SSE-NEXT:    movlhps {{.*#+}} xmm15 = xmm15[0],xmm7[0]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm7[1]
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm13, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm6[1]
+; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm12, %xmm6
+; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm5[0]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm5[1]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm11, %xmm5
+; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm10, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm3[1]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm9, %xmm3
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm8, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 512(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 608(%rdi), %xmm0
+; SSE-NEXT:    movaps 576(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 672(%rdi), %xmm0
+; SSE-NEXT:    movaps 640(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 736(%rdi), %xmm0
+; SSE-NEXT:    movaps 704(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 800(%rdi), %xmm0
+; SSE-NEXT:    movaps 768(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 864(%rdi), %xmm0
+; SSE-NEXT:    movaps 832(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 928(%rdi), %xmm0
+; SSE-NEXT:    movaps 896(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 992(%rdi), %xmm0
+; SSE-NEXT:    movaps 960(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1056(%rdi), %xmm0
+; SSE-NEXT:    movaps 1024(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1120(%rdi), %xmm0
+; SSE-NEXT:    movaps 1088(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1184(%rdi), %xmm0
+; SSE-NEXT:    movaps 1152(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1248(%rdi), %xmm0
+; SSE-NEXT:    movaps 1216(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1312(%rdi), %xmm0
+; SSE-NEXT:    movaps 1280(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1376(%rdi), %xmm0
+; SSE-NEXT:    movaps 1344(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1440(%rdi), %xmm0
+; SSE-NEXT:    movaps 1408(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1504(%rdi), %xmm0
+; SSE-NEXT:    movaps 1472(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1568(%rdi), %xmm0
+; SSE-NEXT:    movaps 1536(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1632(%rdi), %xmm0
+; SSE-NEXT:    movaps 1600(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1696(%rdi), %xmm0
+; SSE-NEXT:    movaps 1664(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1760(%rdi), %xmm0
+; SSE-NEXT:    movaps 1728(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1824(%rdi), %xmm0
+; SSE-NEXT:    movaps 1792(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1888(%rdi), %xmm0
+; SSE-NEXT:    movaps 1856(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1952(%rdi), %xmm0
+; SSE-NEXT:    movaps 1920(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2016(%rdi), %xmm0
+; SSE-NEXT:    movaps 1984(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps (%rdi), %xmm1
+; SSE-NEXT:    movaps 32(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 112(%rdi), %xmm0
+; SSE-NEXT:    movaps 80(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 176(%rdi), %xmm0
+; SSE-NEXT:    movaps 144(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 240(%rdi), %xmm0
+; SSE-NEXT:    movaps 208(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 304(%rdi), %xmm0
+; SSE-NEXT:    movaps 272(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 368(%rdi), %xmm0
+; SSE-NEXT:    movaps 336(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 432(%rdi), %xmm0
+; SSE-NEXT:    movaps 400(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 496(%rdi), %xmm0
+; SSE-NEXT:    movaps 464(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 560(%rdi), %xmm0
+; SSE-NEXT:    movaps 528(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 624(%rdi), %xmm0
+; SSE-NEXT:    movaps 592(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 688(%rdi), %xmm0
+; SSE-NEXT:    movaps 656(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 752(%rdi), %xmm0
+; SSE-NEXT:    movaps 720(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 816(%rdi), %xmm0
+; SSE-NEXT:    movaps 784(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 880(%rdi), %xmm0
+; SSE-NEXT:    movaps 848(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 944(%rdi), %xmm0
+; SSE-NEXT:    movaps 912(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1008(%rdi), %xmm0
+; SSE-NEXT:    movaps 976(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1072(%rdi), %xmm0
+; SSE-NEXT:    movaps 1040(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1136(%rdi), %xmm0
+; SSE-NEXT:    movaps 1104(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1200(%rdi), %xmm0
+; SSE-NEXT:    movaps 1168(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1264(%rdi), %xmm0
+; SSE-NEXT:    movaps 1232(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1328(%rdi), %xmm0
+; SSE-NEXT:    movaps 1296(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1392(%rdi), %xmm0
+; SSE-NEXT:    movaps 1360(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1456(%rdi), %xmm0
+; SSE-NEXT:    movaps 1424(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1520(%rdi), %xmm0
+; SSE-NEXT:    movaps 1488(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1584(%rdi), %xmm0
+; SSE-NEXT:    movaps 1552(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm14
+; SSE-NEXT:    movlhps {{.*#+}} xmm14 = xmm14[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1648(%rdi), %xmm0
+; SSE-NEXT:    movaps 1616(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm11
+; SSE-NEXT:    movlhps {{.*#+}} xmm11 = xmm11[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1712(%rdi), %xmm0
+; SSE-NEXT:    movaps 1680(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1776(%rdi), %xmm0
+; SSE-NEXT:    movaps 1744(%rdi), %xmm12
+; SSE-NEXT:    movaps %xmm12, %xmm9
+; SSE-NEXT:    movlhps {{.*#+}} xmm9 = xmm9[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm0[1]
+; SSE-NEXT:    movaps 1840(%rdi), %xmm0
+; SSE-NEXT:    movaps 1808(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, %xmm3
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm0[1]
+; SSE-NEXT:    movaps 1904(%rdi), %xmm0
+; SSE-NEXT:    movaps 1872(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm0[1]
+; SSE-NEXT:    movaps 1968(%rdi), %xmm0
+; SSE-NEXT:    movaps 1936(%rdi), %xmm6
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm0[1]
+; SSE-NEXT:    movaps 2032(%rdi), %xmm13
+; SSE-NEXT:    movaps 2000(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm13[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm13[1]
+; SSE-NEXT:    movaps 16(%rdi), %xmm13
+; SSE-NEXT:    movaps 48(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm13, %xmm15
+; SSE-NEXT:    movlhps {{.*#+}} xmm15 = xmm15[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm0[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 496(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 480(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 496(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 480(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rdx)
+; SSE-NEXT:    movaps %xmm4, 496(%rcx)
+; SSE-NEXT:    movaps %xmm1, 480(%rcx)
+; SSE-NEXT:    movaps %xmm2, 464(%rcx)
+; SSE-NEXT:    movaps %xmm3, 448(%rcx)
+; SSE-NEXT:    movaps %xmm9, 432(%rcx)
+; SSE-NEXT:    movaps %xmm8, 416(%rcx)
+; SSE-NEXT:    movaps %xmm11, 400(%rcx)
+; SSE-NEXT:    movaps %xmm14, 384(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rcx)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps %xmm15, (%rcx)
+; SSE-NEXT:    movaps %xmm5, 496(%r8)
+; SSE-NEXT:    movaps %xmm6, 480(%r8)
+; SSE-NEXT:    movaps %xmm7, 464(%r8)
+; SSE-NEXT:    movaps %xmm10, 448(%r8)
+; SSE-NEXT:    movaps %xmm12, 432(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r8)
+; SSE-NEXT:    movaps %xmm13, (%r8)
+; SSE-NEXT:    addq $1688, %rsp # imm = 0x698
+; SSE-NEXT:    retq
+;
+; AVX1-ONLY-LABEL: load_i64_stride4_vf64:
+; AVX1-ONLY:       # %bb.0:
+; AVX1-ONLY-NEXT:    subq $2632, %rsp # imm = 0xA48
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm6[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps 1024(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm11[0],xmm10[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm11[1],xmm10[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm6[0],xmm9[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm6[1],xmm9[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm8[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm8[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1312(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1280(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1632(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1600(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1888(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1856(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1824(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1792(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1504(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1472(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1760(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1728(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2016(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1984(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1952(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1920(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 304(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 272(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 560(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 528(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 816(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 784(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1072(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1040(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1328(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1296(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1632(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1600(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1584(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1552(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1888(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1856(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1840(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1808(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2016(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1984(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1968(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1936(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1760(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1728(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1712(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1680(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1504(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1472(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1456(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1424(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1200(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1168(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 944(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vmovaps 912(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm11[0],xmm12[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm9[0],ymm10[0],ymm9[2],ymm10[2]
+; AVX1-ONLY-NEXT:    vmovaps 688(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 656(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm7[0],xmm8[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm14[0],ymm6[0],ymm14[2],ymm6[2]
+; AVX1-ONLY-NEXT:    vmovaps 432(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps 400(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm4[0],xmm5[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm13
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm13[0],ymm2[0],ymm13[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovaps 176(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 144(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm0[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm15[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm13[1],ymm2[1],ymm13[3],ymm2[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm14[1],ymm6[1],ymm14[3],ymm6[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm9[1],ymm10[1],ymm9[3],ymm10[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm8[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm11[1],xmm12[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = xmm5[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 496(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 480(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 432(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 416(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 368(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 352(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 288(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 224(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 160(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 48(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 304(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 240(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 176(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 112(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 464(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 448(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 256(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 384(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 320(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 192(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 128(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, (%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 272(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 400(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 336(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 208(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 144(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 80(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 16(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 224(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 240(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 160(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 176(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 112(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 48(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 480(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 496(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 416(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 432(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 352(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 368(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 288(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 304(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 448(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 464(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 384(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 400(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 320(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 336(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 192(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 208(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, (%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 16(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 80(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 128(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 144(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 256(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 272(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 288(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 352(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 416(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 480(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 448(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 384(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 320(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 256(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, (%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 480(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 416(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 384(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 352(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm6, 320(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm11, 288(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm10, 256(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm9, 224(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 160(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm14, 96(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm13, 32(%r8)
+; AVX1-ONLY-NEXT:    vmovaps %ymm15, (%r8)
+; AVX1-ONLY-NEXT:    addq $2632, %rsp # imm = 0xA48
+; AVX1-ONLY-NEXT:    vzeroupper
+; AVX1-ONLY-NEXT:    retq
+;
+; AVX2-ONLY-LABEL: load_i64_stride4_vf64:
+; AVX2-ONLY:       # %bb.0:
+; AVX2-ONLY-NEXT:    subq $3016, %rsp # imm = 0xBC8
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 608(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vmovaps 576(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1120(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vmovaps 1088(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm8[0],xmm7[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vmovaps 544(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vmovaps 1024(%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm8[1],xmm7[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm11[0],xmm10[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm11[1],xmm10[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 512(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm7[0],xmm9[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm7[1],xmm9[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm6[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm6[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 864(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 832(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 800(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 768(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1312(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1280(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1632(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1600(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1888(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1856(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1824(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1792(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1248(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1152(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1504(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1472(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1440(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1760(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1728(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1696(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1664(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2016(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1984(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1952(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1920(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 736(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 704(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 672(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 640(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 992(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 960(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 928(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 896(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 544(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 512(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 608(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 576(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 800(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 768(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 864(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 832(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1056(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1024(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1120(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1088(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1312(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1280(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1376(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1344(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1568(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1536(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1632(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1600(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1824(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1792(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1888(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1856(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1952(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1920(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2016(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1984(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1696(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1664(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1760(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1728(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1408(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1504(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1472(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1184(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1152(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1216(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 928(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 896(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 992(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vmovaps 960(%rdi), %ymm10
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm11[0],ymm10[2],ymm11[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 672(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovaps 640(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vmovaps 736(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovaps 704(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm13[0],ymm9[0],ymm13[2],ymm9[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %ymm12
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm4[0],ymm2[2],ymm4[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm12[0],ymm5[0],ymm12[2],ymm5[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm3[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %ymm14
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm14[0],ymm3[0],ymm14[2],ymm3[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm15[2,3],ymm6[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm6 = ymm6[1],mem[1],ymm6[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm15[2,3],ymm6[2,3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm14[1],ymm3[1],ymm14[3],ymm3[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm4[1],ymm2[3],ymm4[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm12[1],ymm5[1],ymm12[3],ymm5[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm13[1],ymm9[1],ymm13[3],ymm9[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm11[1],ymm10[3],ymm11[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = ymm4[1],mem[1],ymm4[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm4[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 496(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 480(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 432(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 416(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 368(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 352(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps (%rsp), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 32(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 288(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 224(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 160(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 96(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 48(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 304(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 240(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 176(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 112(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 464(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 448(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 256(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 384(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 320(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 192(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 128(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 64(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, (%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 272(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 400(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 336(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 208(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 144(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 80(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 16(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 224(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 240(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 160(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 176(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 96(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 112(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 32(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 48(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 480(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 496(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 416(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 432(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 352(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 368(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 288(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 304(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 448(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 464(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 384(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 400(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 320(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 336(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 192(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 208(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, (%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 16(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 64(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 80(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 128(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 144(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 256(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 272(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 32(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 96(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 160(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 224(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 288(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 352(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 416(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 480(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 448(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 384(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 320(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 256(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 192(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 128(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 64(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, (%rcx)
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 480(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, 416(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm3, 384(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 352(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm13, 320(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm11, 288(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm10, 256(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm9, 224(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm8, 192(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 160(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm12, 128(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 96(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm15, 64(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm14, 32(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, (%r8)
+; AVX2-ONLY-NEXT:    addq $3016, %rsp # imm = 0xBC8
+; AVX2-ONLY-NEXT:    vzeroupper
+; AVX2-ONLY-NEXT:    retq
+;
+; AVX512-LABEL: load_i64_stride4_vf64:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    subq $2056, %rsp # imm = 0x808
+; AVX512-NEXT:    vmovdqa64 1472(%rdi), %zmm15
+; AVX512-NEXT:    vmovdqa64 1408(%rdi), %zmm0
+; AVX512-NEXT:    vmovdqa64 832(%rdi), %zmm18
+; AVX512-NEXT:    vmovdqa64 768(%rdi), %zmm1
+; AVX512-NEXT:    vmovdqa64 960(%rdi), %zmm22
+; AVX512-NEXT:    vmovdqa64 896(%rdi), %zmm2
+; AVX512-NEXT:    vmovdqa64 576(%rdi), %zmm17
+; AVX512-NEXT:    vmovdqa64 512(%rdi), %zmm3
+; AVX512-NEXT:    vmovdqa64 704(%rdi), %zmm20
+; AVX512-NEXT:    vmovdqa64 640(%rdi), %zmm4
+; AVX512-NEXT:    vmovdqa64 320(%rdi), %zmm27
+; AVX512-NEXT:    vmovdqa64 256(%rdi), %zmm5
+; AVX512-NEXT:    vmovdqa64 448(%rdi), %zmm28
+; AVX512-NEXT:    vmovdqa64 384(%rdi), %zmm6
+; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm11 = [0,4,8,12,0,4,8,12]
+; AVX512-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vmovdqa64 %zmm6, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm28, %zmm11, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm27, %zmm11, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, (%rsp) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm20, %zmm11, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm17, %zmm11, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm22, %zmm11, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm1, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm18, %zmm11, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm11, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm16 = [1,5,9,13,1,5,9,13]
+; AVX512-NEXT:    # zmm16 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vmovdqa64 %zmm6, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm28, %zmm16, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm27, %zmm16, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm22, %zmm16, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm1, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm18, %zmm16, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm20, %zmm16, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm17, %zmm16, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm29 = [2,6,10,14,2,6,10,14]
+; AVX512-NEXT:    # zmm29 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vmovdqa64 %zmm6, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm28, %zmm29, %zmm7
+; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [3,7,11,15,3,7,11,15]
+; AVX512-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT:    vpermt2q %zmm28, %zmm31, %zmm6
+; AVX512-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm6
+; AVX512-NEXT:    vpermt2q %zmm27, %zmm29, %zmm6
+; AVX512-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vpermt2q %zmm27, %zmm31, %zmm5
+; AVX512-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm5
+; AVX512-NEXT:    vpermt2q %zmm22, %zmm29, %zmm5
+; AVX512-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vpermt2q %zmm22, %zmm31, %zmm2
+; AVX512-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm1, %zmm2
+; AVX512-NEXT:    vpermt2q %zmm18, %zmm29, %zmm2
+; AVX512-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vpermt2q %zmm18, %zmm31, %zmm1
+; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm1
+; AVX512-NEXT:    vpermt2q %zmm20, %zmm29, %zmm1
+; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vpermt2q %zmm20, %zmm31, %zmm4
+; AVX512-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm1
+; AVX512-NEXT:    vpermt2q %zmm17, %zmm29, %zmm1
+; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vpermt2q %zmm17, %zmm31, %zmm3
+; AVX512-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm16, %zmm1
+; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm29, %zmm1
+; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm31, %zmm0
+; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 1344(%rdi), %zmm15
+; AVX512-NEXT:    vmovdqa64 1280(%rdi), %zmm20
+; AVX512-NEXT:    vmovdqa64 %zmm20, %zmm0
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm11, %zmm0
+; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm20, %zmm0
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm16, %zmm0
+; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm20, %zmm0
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm29, %zmm0
+; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm31, %zmm20
+; AVX512-NEXT:    vmovdqa64 1216(%rdi), %zmm15
+; AVX512-NEXT:    vmovdqa64 1152(%rdi), %zmm21
+; AVX512-NEXT:    vmovdqa64 %zmm21, %zmm30
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm11, %zmm30
+; AVX512-NEXT:    vmovdqa64 %zmm21, %zmm0
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm16, %zmm0
+; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqa64 %zmm21, %zmm0
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm29, %zmm0
+; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm31, %zmm21
+; AVX512-NEXT:    vmovdqa64 1088(%rdi), %zmm15
+; AVX512-NEXT:    vmovdqa64 1024(%rdi), %zmm17
+; AVX512-NEXT:    vmovdqa64 %zmm17, %zmm26
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm11, %zmm26
+; AVX512-NEXT:    vmovdqa64 %zmm17, %zmm28
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm16, %zmm28
+; AVX512-NEXT:    vmovdqa64 %zmm17, %zmm0
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm29, %zmm0
+; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm31, %zmm17
+; AVX512-NEXT:    vmovdqa64 1984(%rdi), %zmm15
+; AVX512-NEXT:    vmovdqa64 1920(%rdi), %zmm4
+; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm23
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm11, %zmm23
+; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm25
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm16, %zmm25
+; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm27
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm29, %zmm27
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm31, %zmm4
+; AVX512-NEXT:    vmovdqa64 1856(%rdi), %zmm15
+; AVX512-NEXT:    vmovdqa64 1792(%rdi), %zmm5
+; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm19
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm11, %zmm19
+; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm22
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm16, %zmm22
+; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm24
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm29, %zmm24
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm31, %zmm5
+; AVX512-NEXT:    vmovdqa64 1728(%rdi), %zmm15
+; AVX512-NEXT:    vmovdqa64 1664(%rdi), %zmm12
+; AVX512-NEXT:    vmovdqa64 %zmm12, %zmm13
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm11, %zmm13
+; AVX512-NEXT:    vmovdqa64 %zmm12, %zmm14
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm16, %zmm14
+; AVX512-NEXT:    vmovdqa64 %zmm12, %zmm18
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm29, %zmm18
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm31, %zmm12
+; AVX512-NEXT:    vmovdqa64 1600(%rdi), %zmm15
+; AVX512-NEXT:    vmovdqa64 1536(%rdi), %zmm2
+; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm7
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm11, %zmm7
+; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm8
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm16, %zmm8
+; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm10
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm29, %zmm10
+; AVX512-NEXT:    vpermt2q %zmm15, %zmm31, %zmm2
+; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm15
+; AVX512-NEXT:    vmovdqa64 192(%rdi), %zmm0
+; AVX512-NEXT:    vmovdqa64 %zmm15, %zmm3
+; AVX512-NEXT:    vpermt2q %zmm0, %zmm11, %zmm3
+; AVX512-NEXT:    vmovdqa64 %zmm15, %zmm6
+; AVX512-NEXT:    vpermt2q %zmm0, %zmm16, %zmm6
+; AVX512-NEXT:    vmovdqa64 %zmm15, %zmm9
+; AVX512-NEXT:    vpermt2q %zmm0, %zmm29, %zmm9
+; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm15
+; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm1
+; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm0
+; AVX512-NEXT:    vpermi2q %zmm0, %zmm1, %zmm11
+; AVX512-NEXT:    vpermi2q %zmm0, %zmm1, %zmm16
+; AVX512-NEXT:    vpermi2q %zmm0, %zmm1, %zmm29
+; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm1
+; AVX512-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm31 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm31 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm26[0,1,2,3],zmm30[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm30 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm30 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm7 = zmm7[0,1,2,3],zmm13[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm19[0,1,2,3],zmm23[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm11[0,1,2,3],zmm3[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm8[0,1,2,3],zmm14[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm14 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm14 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm19 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm19 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm23 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm23 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm28, %zmm28 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm28 = zmm28[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm22 = zmm22[0,1,2,3],zmm25[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm16[0,1,2,3],zmm6[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm18[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm16 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm16 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm18 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm18 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm25 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm25 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm3 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm3 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm24 = zmm24[0,1,2,3],zmm27[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm29[0,1,2,3],zmm9[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm27 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm27 = zmm8[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm29 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm29 = zmm8[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm8 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm8 = zmm8[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm20, %zmm20 # 64-byte Folded Reload
+; AVX512-NEXT:    # zmm20 = zmm20[0,1,2,3],mem[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm17 = zmm17[0,1,2,3],zmm21[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm4[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm12[4,5,6,7]
+; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm15[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %zmm13, 448(%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm7, 384(%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm30, 320(%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm26, 256(%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm31, 192(%rsi)
+; AVX512-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512-NEXT:    vmovaps %zmm5, 128(%rsi)
+; AVX512-NEXT:    vmovups (%rsp), %zmm5 # 64-byte Reload
+; AVX512-NEXT:    vmovaps %zmm5, 64(%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm11, (%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm22, 448(%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm28, 256(%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm23, 320(%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm19, 128(%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm14, 192(%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm6, (%rdx)
+; AVX512-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512-NEXT:    vmovaps %zmm5, 64(%rdx)
+; AVX512-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512-NEXT:    vmovaps %zmm5, 384(%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm24, 448(%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm0, 256(%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm3, 320(%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm18, 192(%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm9, (%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm16, 64(%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm10, 384(%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm2, 384(%r8)
+; AVX512-NEXT:    vmovdqa64 %zmm4, 448(%r8)
+; AVX512-NEXT:    vmovdqa64 %zmm17, 256(%r8)
+; AVX512-NEXT:    vmovdqa64 %zmm20, 320(%r8)
+; AVX512-NEXT:    vmovdqa64 %zmm8, 128(%r8)
+; AVX512-NEXT:    vmovdqa64 %zmm29, 192(%r8)
+; AVX512-NEXT:    vmovdqa64 %zmm1, (%r8)
+; AVX512-NEXT:    vmovdqa64 %zmm27, 64(%r8)
+; AVX512-NEXT:    addq $2056, %rsp # imm = 0x808
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
+  %wide.vec = load <256 x i64>, ptr %in.vec, align 64
+  %strided.vec0 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <64 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124, i32 128, i32 132, i32 136, i32 140, i32 144, i32 148, i32 152, i32 156, i32 160, i32 164, i32 168, i32 172, i32 176, i32 180, i32 184, i32 188, i32 192, i32 196, i32 200, i32 204, i32 208, i32 212, i32 216, i32 220, i32 224, i32 228, i32 232, i32 236, i32 240, i32 244, i32 248, i32 252>
+  %strided.vec1 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <64 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125, i32 129, i32 133, i32 137, i32 141, i32 145, i32 149, i32 153, i32 157, i32 161, i32 165, i32 169, i32 173, i32 177, i32 181, i32 185, i32 189, i32 193, i32 197, i32 201, i32 205, i32 209, i32 213, i32 217, i32 221, i32 225, i32 229, i32 233, i32 237, i32 241, i32 245, i32 249, i32 253>
+  %strided.vec2 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <64 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126, i32 130, i32 134, i32 138, i32 142, i32 146, i32 150, i32 154, i32 158, i32 162, i32 166, i32 170, i32 174, i32 178, i32 182, i32 186, i32 190, i32 194, i32 198, i32 202, i32 206, i32 210, i32 214, i32 218, i32 222, i32 226, i32 230, i32 234, i32 238, i32 242, i32 246, i32 250, i32 254>
+  %strided.vec3 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <64 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127, i32 131, i32 135, i32 139, i32 143, i32 147, i32 151, i32 155, i32 159, i32 163, i32 167, i32 171, i32 175, i32 179, i32 183, i32 187, i32 191, i32 195, i32 199, i32 203, i32 207, i32 211, i32 215, i32 219, i32 223, i32 227, i32 231, i32 235, i32 239, i32 243, i32 247, i32 251, i32 255>
+  store <64 x i64> %strided.vec0, ptr %out.vec0, align 64
+  store <64 x i64> %strided.vec1, ptr %out.vec1, align 64
+  store <64 x i64> %strided.vec2, ptr %out.vec2, align 64
+  store <64 x i64> %strided.vec3, ptr %out.vec3, align 64
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-5.ll
index 1b1ec07..d096726 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-5.ll
@@ -95,19 +95,16 @@ define void @load_i64_stride5_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <10 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <10 x i64> %wide.vec, <10 x i64> poison, <2 x i32> <i32 0, i32 5>
   %strided.vec1 = shufflevector <10 x i64> %wide.vec, <10 x i64> poison, <2 x i32> <i32 1, i32 6>
   %strided.vec2 = shufflevector <10 x i64> %wide.vec, <10 x i64> poison, <2 x i32> <i32 2, i32 7>
   %strided.vec3 = shufflevector <10 x i64> %wide.vec, <10 x i64> poison, <2 x i32> <i32 3, i32 8>
   %strided.vec4 = shufflevector <10 x i64> %wide.vec, <10 x i64> poison, <2 x i32> <i32 4, i32 9>
-
   store <2 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <2 x i64> %strided.vec3, ptr %out.vec3, align 64
   store <2 x i64> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -343,19 +340,16 @@ define void @load_i64_stride5_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <20 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <20 x i64> %wide.vec, <20 x i64> poison, <4 x i32> <i32 0, i32 5, i32 10, i32 15>
   %strided.vec1 = shufflevector <20 x i64> %wide.vec, <20 x i64> poison, <4 x i32> <i32 1, i32 6, i32 11, i32 16>
   %strided.vec2 = shufflevector <20 x i64> %wide.vec, <20 x i64> poison, <4 x i32> <i32 2, i32 7, i32 12, i32 17>
   %strided.vec3 = shufflevector <20 x i64> %wide.vec, <20 x i64> poison, <4 x i32> <i32 3, i32 8, i32 13, i32 18>
   %strided.vec4 = shufflevector <20 x i64> %wide.vec, <20 x i64> poison, <4 x i32> <i32 4, i32 9, i32 14, i32 19>
-
   store <4 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <4 x i64> %strided.vec3, ptr %out.vec3, align 64
   store <4 x i64> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -701,19 +695,16 @@ define void @load_i64_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <40 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <40 x i64> %wide.vec, <40 x i64> poison, <8 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35>
   %strided.vec1 = shufflevector <40 x i64> %wide.vec, <40 x i64> poison, <8 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36>
   %strided.vec2 = shufflevector <40 x i64> %wide.vec, <40 x i64> poison, <8 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37>
   %strided.vec3 = shufflevector <40 x i64> %wide.vec, <40 x i64> poison, <8 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38>
   %strided.vec4 = shufflevector <40 x i64> %wide.vec, <40 x i64> poison, <8 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39>
-
   store <8 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <8 x i64> %strided.vec3, ptr %out.vec3, align 64
   store <8 x i64> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -1438,19 +1429,16 @@ define void @load_i64_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <80 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <80 x i64> %wide.vec, <80 x i64> poison, <16 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75>
   %strided.vec1 = shufflevector <80 x i64> %wide.vec, <80 x i64> poison, <16 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76>
   %strided.vec2 = shufflevector <80 x i64> %wide.vec, <80 x i64> poison, <16 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77>
   %strided.vec3 = shufflevector <80 x i64> %wide.vec, <80 x i64> poison, <16 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78>
   %strided.vec4 = shufflevector <80 x i64> %wide.vec, <80 x i64> poison, <16 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79>
-
   store <16 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <16 x i64> %strided.vec3, ptr %out.vec3, align 64
   store <16 x i64> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -3032,19 +3020,3497 @@ define void @load_i64_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <160 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <160 x i64> %wide.vec, <160 x i64> poison, <32 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75, i32 80, i32 85, i32 90, i32 95, i32 100, i32 105, i32 110, i32 115, i32 120, i32 125, i32 130, i32 135, i32 140, i32 145, i32 150, i32 155>
   %strided.vec1 = shufflevector <160 x i64> %wide.vec, <160 x i64> poison, <32 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76, i32 81, i32 86, i32 91, i32 96, i32 101, i32 106, i32 111, i32 116, i32 121, i32 126, i32 131, i32 136, i32 141, i32 146, i32 151, i32 156>
   %strided.vec2 = shufflevector <160 x i64> %wide.vec, <160 x i64> poison, <32 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77, i32 82, i32 87, i32 92, i32 97, i32 102, i32 107, i32 112, i32 117, i32 122, i32 127, i32 132, i32 137, i32 142, i32 147, i32 152, i32 157>
   %strided.vec3 = shufflevector <160 x i64> %wide.vec, <160 x i64> poison, <32 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78, i32 83, i32 88, i32 93, i32 98, i32 103, i32 108, i32 113, i32 118, i32 123, i32 128, i32 133, i32 138, i32 143, i32 148, i32 153, i32 158>
   %strided.vec4 = shufflevector <160 x i64> %wide.vec, <160 x i64> poison, <32 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79, i32 84, i32 89, i32 94, i32 99, i32 104, i32 109, i32 114, i32 119, i32 124, i32 129, i32 134, i32 139, i32 144, i32 149, i32 154, i32 159>
-
   store <32 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <32 x i64> %strided.vec3, ptr %out.vec3, align 64
   store <32 x i64> %strided.vec4, ptr %out.vec4, align 64
+  ret void
+}
 
+define void @load_i64_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4) nounwind {
+; SSE-LABEL: load_i64_stride5_vf64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    subq $2200, %rsp # imm = 0x898
+; SSE-NEXT:    movapd 224(%rdi), %xmm0
+; SSE-NEXT:    movapd 144(%rdi), %xmm1
+; SSE-NEXT:    movapd 64(%rdi), %xmm2
+; SSE-NEXT:    movapd 176(%rdi), %xmm3
+; SSE-NEXT:    movapd 96(%rdi), %xmm4
+; SSE-NEXT:    movapd 208(%rdi), %xmm6
+; SSE-NEXT:    movapd 128(%rdi), %xmm7
+; SSE-NEXT:    movapd (%rdi), %xmm9
+; SSE-NEXT:    movapd 16(%rdi), %xmm5
+; SSE-NEXT:    movapd 32(%rdi), %xmm13
+; SSE-NEXT:    movapd 48(%rdi), %xmm8
+; SSE-NEXT:    movapd 160(%rdi), %xmm10
+; SSE-NEXT:    movapd 192(%rdi), %xmm14
+; SSE-NEXT:    movapd 80(%rdi), %xmm11
+; SSE-NEXT:    movapd 112(%rdi), %xmm15
+; SSE-NEXT:    movapd %xmm13, %xmm12
+; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm9[0],xmm12[1]
+; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufpd {{.*#+}} xmm9 = xmm9[1],xmm8[0]
+; SSE-NEXT:    movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm5[0],xmm8[1]
+; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufpd {{.*#+}} xmm5 = xmm5[1],xmm2[0]
+; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm13[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd %xmm15, %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm11[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufpd {{.*#+}} xmm11 = xmm11[1],xmm7[0]
+; SSE-NEXT:    movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm4[0],xmm7[1]
+; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufpd {{.*#+}} xmm4 = xmm4[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm15[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd %xmm14, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm10[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufpd {{.*#+}} xmm10 = xmm10[1],xmm6[0]
+; SSE-NEXT:    movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm3[0],xmm6[1]
+; SSE-NEXT:    movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufpd {{.*#+}} xmm3 = xmm3[1],xmm0[0]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 240(%rdi), %xmm2
+; SSE-NEXT:    movapd 272(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 288(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 256(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 304(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 320(%rdi), %xmm2
+; SSE-NEXT:    movapd 352(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 368(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 336(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 384(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 400(%rdi), %xmm2
+; SSE-NEXT:    movapd 432(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 448(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 416(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 464(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 480(%rdi), %xmm2
+; SSE-NEXT:    movapd 512(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 528(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 496(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 544(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 560(%rdi), %xmm2
+; SSE-NEXT:    movapd 592(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 608(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 576(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 624(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 640(%rdi), %xmm2
+; SSE-NEXT:    movapd 672(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 688(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 656(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 704(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 720(%rdi), %xmm2
+; SSE-NEXT:    movapd 752(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 768(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 736(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 784(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 800(%rdi), %xmm2
+; SSE-NEXT:    movapd 832(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 848(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 816(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 864(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 880(%rdi), %xmm2
+; SSE-NEXT:    movapd 912(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 928(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 896(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 944(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 960(%rdi), %xmm2
+; SSE-NEXT:    movapd 992(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1008(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 976(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1024(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1040(%rdi), %xmm2
+; SSE-NEXT:    movapd 1072(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1088(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1056(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1104(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1120(%rdi), %xmm2
+; SSE-NEXT:    movapd 1152(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1168(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1136(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1184(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1200(%rdi), %xmm2
+; SSE-NEXT:    movapd 1232(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1248(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1216(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1264(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1280(%rdi), %xmm2
+; SSE-NEXT:    movapd 1312(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1328(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1296(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1344(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1360(%rdi), %xmm2
+; SSE-NEXT:    movapd 1392(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1408(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1376(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1424(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1440(%rdi), %xmm2
+; SSE-NEXT:    movapd 1472(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1488(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1456(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1504(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1520(%rdi), %xmm2
+; SSE-NEXT:    movapd 1552(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1568(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1536(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1584(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1600(%rdi), %xmm2
+; SSE-NEXT:    movapd 1632(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1648(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1616(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1664(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1680(%rdi), %xmm2
+; SSE-NEXT:    movapd 1712(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1728(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1696(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1744(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1760(%rdi), %xmm2
+; SSE-NEXT:    movapd 1792(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1808(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1776(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1824(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1840(%rdi), %xmm2
+; SSE-NEXT:    movapd 1872(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1888(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1856(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1904(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1920(%rdi), %xmm2
+; SSE-NEXT:    movapd 1952(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1968(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1936(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1984(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2000(%rdi), %xmm2
+; SSE-NEXT:    movapd 2032(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2048(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2016(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2064(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2080(%rdi), %xmm2
+; SSE-NEXT:    movapd 2112(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2128(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2096(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2144(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2160(%rdi), %xmm14
+; SSE-NEXT:    movapd 2192(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm13
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm14[0],xmm13[1]
+; SSE-NEXT:    movapd 2208(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm14 = xmm14[1],xmm1[0]
+; SSE-NEXT:    movapd 2176(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movapd 2224(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2240(%rdi), %xmm11
+; SSE-NEXT:    movapd 2272(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm9
+; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm11[0],xmm9[1]
+; SSE-NEXT:    movapd 2288(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm11 = xmm11[1],xmm1[0]
+; SSE-NEXT:    movapd 2256(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2304(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2320(%rdi), %xmm7
+; SSE-NEXT:    movapd 2352(%rdi), %xmm1
+; SSE-NEXT:    movapd %xmm1, %xmm5
+; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm7[0],xmm5[1]
+; SSE-NEXT:    movapd 2368(%rdi), %xmm15
+; SSE-NEXT:    shufpd {{.*#+}} xmm7 = xmm7[1],xmm15[0]
+; SSE-NEXT:    movapd 2336(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm2[0],xmm15[1]
+; SSE-NEXT:    movapd 2384(%rdi), %xmm0
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm0[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2400(%rdi), %xmm6
+; SSE-NEXT:    movapd 2432(%rdi), %xmm10
+; SSE-NEXT:    movapd %xmm10, %xmm3
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm6[0],xmm3[1]
+; SSE-NEXT:    movapd 2448(%rdi), %xmm12
+; SSE-NEXT:    shufpd {{.*#+}} xmm6 = xmm6[1],xmm12[0]
+; SSE-NEXT:    movapd 2416(%rdi), %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm1[0],xmm12[1]
+; SSE-NEXT:    movapd 2464(%rdi), %xmm0
+; SSE-NEXT:    shufpd {{.*#+}} xmm1 = xmm1[1],xmm0[0]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm10[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2480(%rdi), %xmm0
+; SSE-NEXT:    movapd 2512(%rdi), %xmm4
+; SSE-NEXT:    movapd %xmm4, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd 2528(%rdi), %xmm8
+; SSE-NEXT:    shufpd {{.*#+}} xmm0 = xmm0[1],xmm8[0]
+; SSE-NEXT:    movapd 2496(%rdi), %xmm10
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm10[0],xmm8[1]
+; SSE-NEXT:    movapd 2544(%rdi), %xmm2
+; SSE-NEXT:    shufpd {{.*#+}} xmm10 = xmm10[1],xmm2[0]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm4[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm1, 496(%rsi)
+; SSE-NEXT:    movapd %xmm3, 480(%rsi)
+; SSE-NEXT:    movapd %xmm5, 464(%rsi)
+; SSE-NEXT:    movapd %xmm9, 448(%rsi)
+; SSE-NEXT:    movapd %xmm13, 432(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 416(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 400(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 384(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 368(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 352(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 336(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 320(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 304(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 288(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 272(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 256(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rsi)
+; SSE-NEXT:    movapd %xmm0, 496(%rdx)
+; SSE-NEXT:    movapd %xmm6, 480(%rdx)
+; SSE-NEXT:    movapd %xmm7, 464(%rdx)
+; SSE-NEXT:    movapd %xmm11, 448(%rdx)
+; SSE-NEXT:    movapd %xmm14, 432(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rdx)
+; SSE-NEXT:    movapd %xmm8, 496(%rcx)
+; SSE-NEXT:    movapd %xmm12, 480(%rcx)
+; SSE-NEXT:    movapd %xmm15, 464(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%rcx)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rcx)
+; SSE-NEXT:    movapd %xmm10, 496(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 480(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r8)
+; SSE-NEXT:    movapd %xmm2, 496(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 480(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r9)
+; SSE-NEXT:    addq $2200, %rsp # imm = 0x898
+; SSE-NEXT:    retq
+;
+; AVX1-ONLY-LABEL: load_i64_stride5_vf64:
+; AVX1-ONLY:       # %bb.0:
+; AVX1-ONLY-NEXT:    subq $3288, %rsp # imm = 0xCD8
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 256(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 192(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 160(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm3 = xmm13[0],xmm3[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm1[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm3[0,1],xmm5[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1216(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1184(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm5[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovapd 1152(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1120(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm7 = xmm4[0],xmm7[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1536(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1504(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm7[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovapd 1472(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vmovapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1440(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm9 = xmm6[0],xmm9[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1856(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1824(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm9[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT:    vmovapd 1792(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vmovapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1760(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm11 = xmm8[0],xmm11[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2176(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2144(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3,4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 2112(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2080(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm10[0,1],xmm12[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm11[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2496(%rdi), %ymm11
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2464(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3,4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 2432(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2400(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm0[0,1],xmm12[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm11[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm11
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm12[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm11[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 416(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 384(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm12[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 352(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 320(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm14 = xmm11[0],xmm0[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm0[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3],ymm14[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1024(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm0[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3],ymm14[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1312(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1280(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm0[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3],ymm14[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1632(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1600(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm0[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3],ymm14[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2016(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1984(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1952(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1920(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm0[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3],ymm14[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2336(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2304(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 2272(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2240(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm0[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3],ymm14[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm14
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm14[0],ymm2[3],ymm14[2]
+; AVX1-ONLY-NEXT:    vmovdqa 208(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm13[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 528(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 848(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm3[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm5[0],ymm0[0],ymm5[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1168(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm4[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm7[0],ymm0[0],ymm7[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1488(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm6[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1888(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm9[0],ymm0[0],ymm9[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1808(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm8[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2208(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2128(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm10[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2528(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2448(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm15[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm12[0],ymm0[0],ymm12[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 368(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm11[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 688(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1008(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1328(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1728(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1648(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vpalignr $8, (%rsp), %xmm1, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2048(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm0[0],ymm3[0],ymm0[3],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1968(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2368(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm0[0],ymm6[0],ymm0[3],ymm6[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2288(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm9[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = xmm6[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 176(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm7[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm11[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 496(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 816(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 1136(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 1456(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1856(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 1776(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2176(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 2096(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2496(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 2416(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2336(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm0[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX1-ONLY-NEXT:    vmovdqa 2256(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm9 = xmm0[0,1,2,3],xmm9[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2016(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm0[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovdqa 1936(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm0[0,1,2,3],xmm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovdqa 1616(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm2[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovdqa 1296(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm4[0,1,2,3],xmm5[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 976(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = xmm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = xmm3[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm8
+; AVX1-ONLY-NEXT:    vmovaps 656(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm9 = xmm5[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm9 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
+; AVX1-ONLY-NEXT:    vmovaps 336(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm10[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm12[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm9[0,1],xmm14[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX1-ONLY-NEXT:    vmovdqa 16(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm15 = xmm14[0,1,2,3],xmm15[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1,2,3],ymm12[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
+; AVX1-ONLY-NEXT:    vmovapd 128(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vmovupd %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm9 = ymm9[0],ymm12[0],ymm9[3],ymm12[2]
+; AVX1-ONLY-NEXT:    vmovdqa 64(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm12 = xmm14[8,9,10,11,12,13,14,15],xmm12[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm12[0,1],ymm9[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vmovapd 288(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm9[0],ymm6[3],ymm9[2]
+; AVX1-ONLY-NEXT:    vmovdqa 224(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm7 = xmm7[8,9,10,11,12,13,14,15],xmm9[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vmovapd 448(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[3],ymm7[2]
+; AVX1-ONLY-NEXT:    vmovdqa 384(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm7 = xmm10[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vmovapd 608(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[3],ymm7[2]
+; AVX1-ONLY-NEXT:    vmovdqa 544(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vmovapd 768(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm3 = ymm3[0],ymm6[0],ymm3[3],ymm6[2]
+; AVX1-ONLY-NEXT:    vmovdqa 704(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm5 = xmm5[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rsp), %ymm0, %ymm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 928(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm3 = ymm3[0],ymm5[0],ymm3[3],ymm5[2]
+; AVX1-ONLY-NEXT:    vmovdqa 864(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = mem[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 1088(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[3],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1024(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 1248(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1184(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 1408(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1344(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm4[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 1568(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm14[0],ymm0[3],ymm14[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1504(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 1728(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm12[0],ymm0[3],ymm12[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1664(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 1888(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm10[0],ymm0[3],ymm10[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1824(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm9[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 2048(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm8[0],ymm0[3],ymm8[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1984(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 2208(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[3],ymm6[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2144(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 2368(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[3],ymm4[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2304(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 2528(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[3],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2464(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT:    vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = mem[0,1,2,3],xmm13[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm13[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT:    vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = mem[0,1,2,3],xmm11[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm10[3]
+; AVX1-ONLY-NEXT:    vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm9 = mem[0,1,2,3],xmm9[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT:    vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = mem[0,1,2,3],xmm7[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT:    vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = mem[0,1,2,3],xmm5[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = mem[0,1,2,3],xmm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT:    vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0,1,2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 448(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 384(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 320(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 256(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 128(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 480(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 416(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 352(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 288(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 160(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 448(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 384(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 320(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 256(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 128(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 480(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 416(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 352(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 288(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 160(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 256(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 320(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 384(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 448(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 480(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 416(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 352(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 288(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 480(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 448(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 416(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 384(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 352(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 320(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 288(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 256(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%r8)
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 160(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 128(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 32(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%r8)
+; AVX1-ONLY-NEXT:    vmovapd %ymm0, 480(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm3, 448(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm5, 416(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm7, 384(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm9, 352(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm11, 320(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm13, 288(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm15, 256(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%r9)
+; AVX1-ONLY-NEXT:    addq $3288, %rsp # imm = 0xCD8
+; AVX1-ONLY-NEXT:    vzeroupper
+; AVX1-ONLY-NEXT:    retq
+;
+; AVX2-ONLY-LABEL: load_i64_stride5_vf64:
+; AVX2-ONLY:       # %bb.0:
+; AVX2-ONLY-NEXT:    subq $3240, %rsp # imm = 0xCA8
+; AVX2-ONLY-NEXT:    vmovdqa 896(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 864(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 576(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 544(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 512(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 480(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 832(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 800(%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm11[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1216(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1184(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1152(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1120(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm12[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1536(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1504(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1472(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1440(%rdi), %xmm13
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1856(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1824(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1792(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1760(%rdi), %xmm14
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm14[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2176(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2144(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2112(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2080(%rdi), %xmm15
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2496(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2464(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 2432(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2400(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 416(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 384(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 352(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 320(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm9[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 736(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 704(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 672(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 640(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm1[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1056(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1024(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 992(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 960(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm2[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1376(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1344(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1312(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1280(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm3[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1696(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1664(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1632(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1600(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm6 = xmm4[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2016(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1984(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1952(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1920(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm7 = xmm5[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2336(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2304(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm7 = ymm6[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2272(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2240(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm8 = xmm6[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 208(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 288(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 528(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm7 = xmm10[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 608(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 848(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm7 = xmm11[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 928(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr $8, (%rsp), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm0, %ymm10
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1168(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm7 = xmm12[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1248(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1488(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm7 = xmm13[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1568(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1808(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm7 = xmm14[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1888(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2128(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm7 = xmm15[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2208(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm0, %ymm15
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2448(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2528(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 128(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm0, %ymm12
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 368(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm9[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 448(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = mem[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm8, %ymm11
+; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 688(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 768(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm7, %ymm9
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1008(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm2[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1088(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm2, %ymm8
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1328(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm3[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1408(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm2, %ymm7
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1648(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm4[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1728(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa %ymm2, %ymm4
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1968(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm5[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2048(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2288(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm6[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2368(%rdi), %ymm14
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm14[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm14[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 576(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 800(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 896(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1120(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1760(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 1856(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2080(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 2176(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2400(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 2496(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2240(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 2336(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1920(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 2016(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1600(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 1696(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1280(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 960(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 640(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 736(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm12[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm12[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 64(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 224(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm11[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm11[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 384(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm12[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 544(%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 704(%rdi), %xmm13
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm10[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm10[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 864(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 1024(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm9[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 1184(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 1344(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu (%rsp), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 1504(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 1664(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 1824(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 1984(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm15[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm15[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 2144(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm14[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm14[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 2304(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm15 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm15 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovdqa 2464(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm14 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm12 = mem[0,1],xmm12[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm12 = mem[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm11 = mem[0,1],xmm11[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm12 = mem[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm13 = mem[0,1],xmm13[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm12 = mem[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm10 = mem[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm12 = mem[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm9 = mem[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm15 = ymm9[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm9 = mem[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm8 = mem[0,1],xmm8[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm12 = ymm8[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm8 = mem[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = mem[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm14 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu (%rsp), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = mem[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm6 = mem[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm9 = ymm6[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm6 = mem[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm5 = mem[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm6 = mem[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm4 = mem[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm4[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = mem[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = mem[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = mem[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 480(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 416(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 480(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 416(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 480(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 416(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 480(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 416(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%r8)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm0, 480(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm3, 448(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm7, 416(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm4, 384(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm6, 352(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm5, 320(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm9, 288(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm14, 256(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm12, 224(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm15, 192(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm10, 160(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm13, 128(%r9)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm11, 96(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 64(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 32(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, (%r9)
+; AVX2-ONLY-NEXT:    addq $3240, %rsp # imm = 0xCA8
+; AVX2-ONLY-NEXT:    vzeroupper
+; AVX2-ONLY-NEXT:    retq
+;
+; AVX512F-LABEL: load_i64_stride5_vf64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    subq $3336, %rsp # imm = 0xD08
+; AVX512F-NEXT:    vmovdqa64 1728(%rdi), %zmm16
+; AVX512F-NEXT:    vmovdqa64 1792(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqa64 1408(%rdi), %zmm15
+; AVX512F-NEXT:    vmovdqa64 1088(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqa64 1152(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqa64 768(%rdi), %zmm19
+; AVX512F-NEXT:    vmovdqa64 832(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqa64 448(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 512(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 128(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa64 192(%rdi), %zmm7
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [12,1,6,0,12,1,6,0]
+; AVX512F-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm8
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm9, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm8
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm9, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm8
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm9, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm8
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm9, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm10
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm8
+; AVX512F-NEXT:    vpermt2q %zmm16, %zmm9, %zmm10
+; AVX512F-NEXT:    vmovdqu64 %zmm10, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [5,10,15,0,5,10,15,0]
+; AVX512F-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm10, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm10, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm10, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm19, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm10, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm10, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm11 = [6,11,0,1,6,11,0,1]
+; AVX512F-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm11, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm11, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm11, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm19, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm11, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm11, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm12 = [7,12,0,2,7,12,0,2]
+; AVX512F-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm12, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [0,5,0,11,0,5,0,11]
+; AVX512F-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm3, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm12, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm3, %zmm6
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm12, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm3, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm19, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm12, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm3, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm12, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm16, %zmm3, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1472(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm18
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm9, %zmm18
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm10, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm11, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm12, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm3, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2048(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 2112(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm19
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm9, %zmm19
+; AVX512F-NEXT:    vmovdqa64 2368(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa64 2432(%rdi), %zmm6
+; AVX512F-NEXT:    vpermi2q %zmm1, %zmm6, %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm10, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm6, %zmm1, %zmm10
+; AVX512F-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm11, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm6, %zmm1, %zmm11
+; AVX512F-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm12, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm6, %zmm1, %zmm12
+; AVX512F-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm3, %zmm6
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm3, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 (%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 64(%rdi), %zmm31
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,10,15]
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm5
+; AVX512F-NEXT:    vpermt2q %zmm31, %zmm1, %zmm5
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} ymm28 = <1,6,11,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm31, %zmm28, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm8 = <2,7,12,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm31, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm11 = <11,0,5,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm11, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm15 = <12,1,6,u>
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm15, %zmm31
+; AVX512F-NEXT:    vmovdqa64 384(%rdi), %zmm27
+; AVX512F-NEXT:    vmovdqa64 320(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm17
+; AVX512F-NEXT:    vpermt2q %zmm27, %zmm1, %zmm17
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm27, %zmm28, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm27, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm11, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm15, %zmm27
+; AVX512F-NEXT:    vmovdqa64 1024(%rdi), %zmm26
+; AVX512F-NEXT:    vmovdqa64 960(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm1, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm28, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm26, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm11, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm15, %zmm26
+; AVX512F-NEXT:    vmovdqa64 704(%rdi), %zmm30
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm14
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm1, %zmm14
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm28, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm11, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm15, %zmm30
+; AVX512F-NEXT:    vmovdqa64 1664(%rdi), %zmm24
+; AVX512F-NEXT:    vmovdqa64 1600(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm12
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm1, %zmm12
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm28, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm25
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm8, %zmm25
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm11, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm15, %zmm24
+; AVX512F-NEXT:    vmovdqa64 1344(%rdi), %zmm13
+; AVX512F-NEXT:    vmovdqa64 1280(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm13, %zmm1, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm6
+; AVX512F-NEXT:    vpermt2q %zmm13, %zmm28, %zmm6
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm22
+; AVX512F-NEXT:    vpermt2q %zmm13, %zmm8, %zmm22
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm23
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm11, %zmm23
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm15, %zmm13
+; AVX512F-NEXT:    vmovdqa64 1984(%rdi), %zmm10
+; AVX512F-NEXT:    vmovdqa64 1920(%rdi), %zmm9
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm29
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm1, %zmm29
+; AVX512F-NEXT:    vmovdqa64 2304(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqa64 2240(%rdi), %zmm0
+; AVX512F-NEXT:    vpermi2q %zmm7, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm16
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm28, %zmm16
+; AVX512F-NEXT:    vpermi2q %zmm7, %zmm0, %zmm28
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm20
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm8, %zmm20
+; AVX512F-NEXT:    vpermi2q %zmm7, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm21
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm11, %zmm21
+; AVX512F-NEXT:    vpermi2q %zmm0, %zmm7, %zmm11
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm15, %zmm7
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm15, %zmm10
+; AVX512F-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm5, %zmm9 # 64-byte Folded Reload
+; AVX512F-NEXT:    # zmm9 = zmm5[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm17, %zmm2 # 64-byte Folded Reload
+; AVX512F-NEXT:    # zmm2 = zmm17[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm14, %zmm14 # 64-byte Folded Reload
+; AVX512F-NEXT:    # zmm14 = zmm14[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm3 # 64-byte Folded Reload
+; AVX512F-NEXT:    # zmm3 = zmm3[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm18[4,5,6,7]
+; AVX512F-NEXT:    vshufi64x2 $228, (%rsp), %zmm12, %zmm5 # 64-byte Folded Reload
+; AVX512F-NEXT:    # zmm5 = zmm12[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm29[0,1,2,3],zmm19[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
+; AVX512F-NEXT:    # zmm0 = zmm1[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 256(%rdi), %zmm29
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,4,5,6,11]
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm0, %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %zmm15
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 896(%rdi), %zmm18
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm0, %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1216(%rdi), %zmm14
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1536(%rdi), %zmm17
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm0, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1856(%rdi), %zmm12
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2176(%rdi), %zmm19
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2496(%rdi), %zmm9
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    movb $7, %al
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm0 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm16 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm4 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm6 {%k1}
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm28 = [0,1,2,3,4,5,6,12]
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm28, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm28, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm28, %zmm16
+; AVX512F-NEXT:    vmovdqu64 %zmm16, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm28, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm28, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm28, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm28, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm16
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm28, %zmm6
+; AVX512F-NEXT:    movb $56, %al
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm25 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm22 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k1}
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,4,5,8,13]
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm0, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm0, %zmm20
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm0, %zmm25
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm0, %zmm22
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm28 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm23 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k1}
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,4,5,9,14]
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm0, %zmm28
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm0, %zmm21
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm0, %zmm5
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm0, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm0, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm0, %zmm23
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm11
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm31 {%k1}
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,4,5,10,15]
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm0, %zmm31
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm27 {%k1}
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm0, %zmm27
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm26 {%k1}
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm0, %zmm26
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm30 {%k1}
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm0, %zmm30
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm24 {%k1}
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm0, %zmm24
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm13 {%k1}
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm0, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm7 {%k1}
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm7
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm10 {%k1}
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm0, %zmm10
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 448(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 384(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 320(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 256(%rsi)
+; AVX512F-NEXT:    vmovups (%rsp), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm6, 448(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm16, 256(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 320(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 384(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm8, 448(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm22, 256(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm25, 320(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm20, 384(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm11, 448(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm23, 256(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm1, 320(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm3, 128(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm2, 192(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm28, (%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm5, 64(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm21, 384(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm10, 384(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm7, 448(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm13, 256(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm24, 320(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm30, 128(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm26, 192(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm31, (%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm27, 64(%r9)
+; AVX512F-NEXT:    addq $3336, %rsp # imm = 0xD08
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: load_i64_stride5_vf64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    subq $3336, %rsp # imm = 0xD08
+; AVX512BW-NEXT:    vmovdqa64 1728(%rdi), %zmm16
+; AVX512BW-NEXT:    vmovdqa64 1792(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqa64 1408(%rdi), %zmm15
+; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm19
+; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm5
+; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm7
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [12,1,6,0,12,1,6,0]
+; AVX512BW-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm8
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm9, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm8
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm8
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm9, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm8
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm9, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm10
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm8
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm9, %zmm10
+; AVX512BW-NEXT:    vmovdqu64 %zmm10, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [5,10,15,0,5,10,15,0]
+; AVX512BW-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm10, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm10, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm10, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm10, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm10, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm11 = [6,11,0,1,6,11,0,1]
+; AVX512BW-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm11, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm11, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm11, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm11, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm11, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm12 = [7,12,0,2,7,12,0,2]
+; AVX512BW-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm12, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [0,5,0,11,0,5,0,11]
+; AVX512BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm3, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm12, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm3, %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm12, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm3, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm12, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm3, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm12, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm3, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1472(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm18
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm9, %zmm18
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm10, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm11, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm12, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2048(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 2112(%rdi), %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm19
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm19
+; AVX512BW-NEXT:    vmovdqa64 2368(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa64 2432(%rdi), %zmm6
+; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm6, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm10, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm6, %zmm1, %zmm10
+; AVX512BW-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm11, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm6, %zmm1, %zmm11
+; AVX512BW-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm12, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm6, %zmm1, %zmm12
+; AVX512BW-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm3, %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm3, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm31
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,10,15]
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm5
+; AVX512BW-NEXT:    vpermt2q %zmm31, %zmm1, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} ymm28 = <1,6,11,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm31, %zmm28, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm8 = <2,7,12,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm31, %zmm8, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm11 = <11,0,5,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm11, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm15 = <12,1,6,u>
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm31
+; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm27
+; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm17
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm1, %zmm17
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm28, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm8, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm11, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm27
+; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %zmm26
+; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm1, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm28, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm8, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm11, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm26
+; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm30
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm14
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm1, %zmm14
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm28, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm8, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm11, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm30
+; AVX512BW-NEXT:    vmovdqa64 1664(%rdi), %zmm24
+; AVX512BW-NEXT:    vmovdqa64 1600(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm12
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm1, %zmm12
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm28, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm25
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm8, %zmm25
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm11, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm24
+; AVX512BW-NEXT:    vmovdqa64 1344(%rdi), %zmm13
+; AVX512BW-NEXT:    vmovdqa64 1280(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm1, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm6
+; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm28, %zmm6
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm22
+; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm8, %zmm22
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm23
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm11, %zmm23
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm13
+; AVX512BW-NEXT:    vmovdqa64 1984(%rdi), %zmm10
+; AVX512BW-NEXT:    vmovdqa64 1920(%rdi), %zmm9
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm29
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm1, %zmm29
+; AVX512BW-NEXT:    vmovdqa64 2304(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqa64 2240(%rdi), %zmm0
+; AVX512BW-NEXT:    vpermi2q %zmm7, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm16
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm28, %zmm16
+; AVX512BW-NEXT:    vpermi2q %zmm7, %zmm0, %zmm28
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm20
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm8, %zmm20
+; AVX512BW-NEXT:    vpermi2q %zmm7, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm21
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm11, %zmm21
+; AVX512BW-NEXT:    vpermi2q %zmm0, %zmm7, %zmm11
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm7
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm15, %zmm10
+; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm5, %zmm9 # 64-byte Folded Reload
+; AVX512BW-NEXT:    # zmm9 = zmm5[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm17, %zmm2 # 64-byte Folded Reload
+; AVX512BW-NEXT:    # zmm2 = zmm17[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm14, %zmm14 # 64-byte Folded Reload
+; AVX512BW-NEXT:    # zmm14 = zmm14[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm3 # 64-byte Folded Reload
+; AVX512BW-NEXT:    # zmm3 = zmm3[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm18[4,5,6,7]
+; AVX512BW-NEXT:    vshufi64x2 $228, (%rsp), %zmm12, %zmm5 # 64-byte Folded Reload
+; AVX512BW-NEXT:    # zmm5 = zmm12[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm29[0,1,2,3],zmm19[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
+; AVX512BW-NEXT:    # zmm0 = zmm1[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm29
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,4,5,6,11]
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm0, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm15
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm18
+; AVX512BW-NEXT:    vpermt2q %zmm18, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %zmm14
+; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1536(%rdi), %zmm17
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1856(%rdi), %zmm12
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2176(%rdi), %zmm19
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2496(%rdi), %zmm9
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    movb $7, %al
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm16 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm4 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm6 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm28 = [0,1,2,3,4,5,6,12]
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm28, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm28, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm28, %zmm16
+; AVX512BW-NEXT:    vmovdqu64 %zmm16, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm28, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm18, %zmm28, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm28, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm28, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm16
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm28, %zmm6
+; AVX512BW-NEXT:    movb $56, %al
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm25 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm22 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,4,5,8,13]
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm0, %zmm20
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm18, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm0, %zmm25
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm0, %zmm22
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm28 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm23 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,4,5,9,14]
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm0, %zmm28
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm0, %zmm21
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm0, %zmm5
+; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm18, %zmm0, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm0, %zmm23
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm0, %zmm11
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm31 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,4,5,10,15]
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm0, %zmm31
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm27 {%k1}
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm0, %zmm27
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm26 {%k1}
+; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm0, %zmm26
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm30 {%k1}
+; AVX512BW-NEXT:    vpermt2q %zmm18, %zmm0, %zmm30
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm24 {%k1}
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm0, %zmm24
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm13 {%k1}
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm0, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm7 {%k1}
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm0, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm10 {%k1}
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm0, %zmm10
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 448(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 384(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 320(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 256(%rsi)
+; AVX512BW-NEXT:    vmovups (%rsp), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, 448(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, 256(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 320(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 384(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, 448(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, 256(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, 320(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, 384(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, 448(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, 256(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, 320(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, 128(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, 192(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, (%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, 64(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, 384(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, 384(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, 448(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, 256(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, 320(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, 128(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, 192(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, (%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, 64(%r9)
+; AVX512BW-NEXT:    addq $3336, %rsp # imm = 0xD08
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %wide.vec = load <320 x i64>, ptr %in.vec, align 64
+  %strided.vec0 = shufflevector <320 x i64> %wide.vec, <320 x i64> poison, <64 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75, i32 80, i32 85, i32 90, i32 95, i32 100, i32 105, i32 110, i32 115, i32 120, i32 125, i32 130, i32 135, i32 140, i32 145, i32 150, i32 155, i32 160, i32 165, i32 170, i32 175, i32 180, i32 185, i32 190, i32 195, i32 200, i32 205, i32 210, i32 215, i32 220, i32 225, i32 230, i32 235, i32 240, i32 245, i32 250, i32 255, i32 260, i32 265, i32 270, i32 275, i32 280, i32 285, i32 290, i32 295, i32 300, i32 305, i32 310, i32 315>
+  %strided.vec1 = shufflevector <320 x i64> %wide.vec, <320 x i64> poison, <64 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76, i32 81, i32 86, i32 91, i32 96, i32 101, i32 106, i32 111, i32 116, i32 121, i32 126, i32 131, i32 136, i32 141, i32 146, i32 151, i32 156, i32 161, i32 166, i32 171, i32 176, i32 181, i32 186, i32 191, i32 196, i32 201, i32 206, i32 211, i32 216, i32 221, i32 226, i32 231, i32 236, i32 241, i32 246, i32 251, i32 256, i32 261, i32 266, i32 271, i32 276, i32 281, i32 286, i32 291, i32 296, i32 301, i32 306, i32 311, i32 316>
+  %strided.vec2 = shufflevector <320 x i64> %wide.vec, <320 x i64> poison, <64 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77, i32 82, i32 87, i32 92, i32 97, i32 102, i32 107, i32 112, i32 117, i32 122, i32 127, i32 132, i32 137, i32 142, i32 147, i32 152, i32 157, i32 162, i32 167, i32 172, i32 177, i32 182, i32 187, i32 192, i32 197, i32 202, i32 207, i32 212, i32 217, i32 222, i32 227, i32 232, i32 237, i32 242, i32 247, i32 252, i32 257, i32 262, i32 267, i32 272, i32 277, i32 282, i32 287, i32 292, i32 297, i32 302, i32 307, i32 312, i32 317>
+  %strided.vec3 = shufflevector <320 x i64> %wide.vec, <320 x i64> poison, <64 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78, i32 83, i32 88, i32 93, i32 98, i32 103, i32 108, i32 113, i32 118, i32 123, i32 128, i32 133, i32 138, i32 143, i32 148, i32 153, i32 158, i32 163, i32 168, i32 173, i32 178, i32 183, i32 188, i32 193, i32 198, i32 203, i32 208, i32 213, i32 218, i32 223, i32 228, i32 233, i32 238, i32 243, i32 248, i32 253, i32 258, i32 263, i32 268, i32 273, i32 278, i32 283, i32 288, i32 293, i32 298, i32 303, i32 308, i32 313, i32 318>
+  %strided.vec4 = shufflevector <320 x i64> %wide.vec, <320 x i64> poison, <64 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79, i32 84, i32 89, i32 94, i32 99, i32 104, i32 109, i32 114, i32 119, i32 124, i32 129, i32 134, i32 139, i32 144, i32 149, i32 154, i32 159, i32 164, i32 169, i32 174, i32 179, i32 184, i32 189, i32 194, i32 199, i32 204, i32 209, i32 214, i32 219, i32 224, i32 229, i32 234, i32 239, i32 244, i32 249, i32 254, i32 259, i32 264, i32 269, i32 274, i32 279, i32 284, i32 289, i32 294, i32 299, i32 304, i32 309, i32 314, i32 319>
+  store <64 x i64> %strided.vec0, ptr %out.vec0, align 64
+  store <64 x i64> %strided.vec1, ptr %out.vec1, align 64
+  store <64 x i64> %strided.vec2, ptr %out.vec2, align 64
+  store <64 x i64> %strided.vec3, ptr %out.vec3, align 64
+  store <64 x i64> %strided.vec4, ptr %out.vec4, align 64
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-6.ll
index 44b432f..3eca48f 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-6.ll
@@ -65,21 +65,18 @@ define void @load_i64_stride6_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX-NEXT:    vmovaps %xmm2, (%rax)
 ; AVX-NEXT:    retq
   %wide.vec = load <12 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <12 x i64> %wide.vec, <12 x i64> poison, <2 x i32> <i32 0, i32 6>
   %strided.vec1 = shufflevector <12 x i64> %wide.vec, <12 x i64> poison, <2 x i32> <i32 1, i32 7>
   %strided.vec2 = shufflevector <12 x i64> %wide.vec, <12 x i64> poison, <2 x i32> <i32 2, i32 8>
   %strided.vec3 = shufflevector <12 x i64> %wide.vec, <12 x i64> poison, <2 x i32> <i32 3, i32 9>
   %strided.vec4 = shufflevector <12 x i64> %wide.vec, <12 x i64> poison, <2 x i32> <i32 4, i32 10>
   %strided.vec5 = shufflevector <12 x i64> %wide.vec, <12 x i64> poison, <2 x i32> <i32 5, i32 11>
-
   store <2 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <2 x i64> %strided.vec3, ptr %out.vec3, align 64
   store <2 x i64> %strided.vec4, ptr %out.vec4, align 64
   store <2 x i64> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -384,21 +381,18 @@ define void @load_i64_stride6_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <24 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <24 x i64> %wide.vec, <24 x i64> poison, <4 x i32> <i32 0, i32 6, i32 12, i32 18>
   %strided.vec1 = shufflevector <24 x i64> %wide.vec, <24 x i64> poison, <4 x i32> <i32 1, i32 7, i32 13, i32 19>
   %strided.vec2 = shufflevector <24 x i64> %wide.vec, <24 x i64> poison, <4 x i32> <i32 2, i32 8, i32 14, i32 20>
   %strided.vec3 = shufflevector <24 x i64> %wide.vec, <24 x i64> poison, <4 x i32> <i32 3, i32 9, i32 15, i32 21>
   %strided.vec4 = shufflevector <24 x i64> %wide.vec, <24 x i64> poison, <4 x i32> <i32 4, i32 10, i32 16, i32 22>
   %strided.vec5 = shufflevector <24 x i64> %wide.vec, <24 x i64> poison, <4 x i32> <i32 5, i32 11, i32 17, i32 23>
-
   store <4 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <4 x i64> %strided.vec3, ptr %out.vec3, align 64
   store <4 x i64> %strided.vec4, ptr %out.vec4, align 64
   store <4 x i64> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -853,21 +847,18 @@ define void @load_i64_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <48 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <48 x i64> %wide.vec, <48 x i64> poison, <8 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42>
   %strided.vec1 = shufflevector <48 x i64> %wide.vec, <48 x i64> poison, <8 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43>
   %strided.vec2 = shufflevector <48 x i64> %wide.vec, <48 x i64> poison, <8 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44>
   %strided.vec3 = shufflevector <48 x i64> %wide.vec, <48 x i64> poison, <8 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45>
   %strided.vec4 = shufflevector <48 x i64> %wide.vec, <48 x i64> poison, <8 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46>
   %strided.vec5 = shufflevector <48 x i64> %wide.vec, <48 x i64> poison, <8 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47>
-
   store <8 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <8 x i64> %strided.vec3, ptr %out.vec3, align 64
   store <8 x i64> %strided.vec4, ptr %out.vec4, align 64
   store <8 x i64> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -1815,21 +1806,18 @@ define void @load_i64_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <96 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <96 x i64> %wide.vec, <96 x i64> poison, <16 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90>
   %strided.vec1 = shufflevector <96 x i64> %wide.vec, <96 x i64> poison, <16 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91>
   %strided.vec2 = shufflevector <96 x i64> %wide.vec, <96 x i64> poison, <16 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92>
   %strided.vec3 = shufflevector <96 x i64> %wide.vec, <96 x i64> poison, <16 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93>
   %strided.vec4 = shufflevector <96 x i64> %wide.vec, <96 x i64> poison, <16 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94>
   %strided.vec5 = shufflevector <96 x i64> %wide.vec, <96 x i64> poison, <16 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95>
-
   store <16 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <16 x i64> %strided.vec3, ptr %out.vec3, align 64
   store <16 x i64> %strided.vec4, ptr %out.vec4, align 64
   store <16 x i64> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -4016,21 +4004,4702 @@ define void @load_i64_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <192 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <192 x i64> %wide.vec, <192 x i64> poison, <32 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90, i32 96, i32 102, i32 108, i32 114, i32 120, i32 126, i32 132, i32 138, i32 144, i32 150, i32 156, i32 162, i32 168, i32 174, i32 180, i32 186>
   %strided.vec1 = shufflevector <192 x i64> %wide.vec, <192 x i64> poison, <32 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91, i32 97, i32 103, i32 109, i32 115, i32 121, i32 127, i32 133, i32 139, i32 145, i32 151, i32 157, i32 163, i32 169, i32 175, i32 181, i32 187>
   %strided.vec2 = shufflevector <192 x i64> %wide.vec, <192 x i64> poison, <32 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92, i32 98, i32 104, i32 110, i32 116, i32 122, i32 128, i32 134, i32 140, i32 146, i32 152, i32 158, i32 164, i32 170, i32 176, i32 182, i32 188>
   %strided.vec3 = shufflevector <192 x i64> %wide.vec, <192 x i64> poison, <32 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93, i32 99, i32 105, i32 111, i32 117, i32 123, i32 129, i32 135, i32 141, i32 147, i32 153, i32 159, i32 165, i32 171, i32 177, i32 183, i32 189>
   %strided.vec4 = shufflevector <192 x i64> %wide.vec, <192 x i64> poison, <32 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94, i32 100, i32 106, i32 112, i32 118, i32 124, i32 130, i32 136, i32 142, i32 148, i32 154, i32 160, i32 166, i32 172, i32 178, i32 184, i32 190>
   %strided.vec5 = shufflevector <192 x i64> %wide.vec, <192 x i64> poison, <32 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95, i32 101, i32 107, i32 113, i32 119, i32 125, i32 131, i32 137, i32 143, i32 149, i32 155, i32 161, i32 167, i32 173, i32 179, i32 185, i32 191>
-
   store <32 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i64> %strided.vec2, ptr %out.vec2, align 64
   store <32 x i64> %strided.vec3, ptr %out.vec3, align 64
   store <32 x i64> %strided.vec4, ptr %out.vec4, align 64
   store <32 x i64> %strided.vec5, ptr %out.vec5, align 64
+  ret void
+}
 
+define void @load_i64_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
+; SSE-LABEL: load_i64_stride6_vf64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    subq $2712, %rsp # imm = 0xA98
+; SSE-NEXT:    movaps 816(%rdi), %xmm0
+; SSE-NEXT:    movaps 720(%rdi), %xmm1
+; SSE-NEXT:    movaps 672(%rdi), %xmm8
+; SSE-NEXT:    movaps 624(%rdi), %xmm2
+; SSE-NEXT:    movaps 576(%rdi), %xmm9
+; SSE-NEXT:    movaps 528(%rdi), %xmm3
+; SSE-NEXT:    movaps 480(%rdi), %xmm10
+; SSE-NEXT:    movaps 432(%rdi), %xmm4
+; SSE-NEXT:    movaps 384(%rdi), %xmm11
+; SSE-NEXT:    movaps 336(%rdi), %xmm5
+; SSE-NEXT:    movaps 288(%rdi), %xmm12
+; SSE-NEXT:    movaps 240(%rdi), %xmm6
+; SSE-NEXT:    movaps 192(%rdi), %xmm13
+; SSE-NEXT:    movaps 144(%rdi), %xmm7
+; SSE-NEXT:    movaps 96(%rdi), %xmm14
+; SSE-NEXT:    movaps %xmm14, %xmm15
+; SSE-NEXT:    movlhps {{.*#+}} xmm15 = xmm15[0],xmm7[0]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm7[1]
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm13, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm6[1]
+; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm12, %xmm6
+; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm5[0]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm5[1]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm11, %xmm5
+; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm10, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm3[1]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm9, %xmm3
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm8, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 768(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 912(%rdi), %xmm0
+; SSE-NEXT:    movaps 864(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1008(%rdi), %xmm0
+; SSE-NEXT:    movaps 960(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1104(%rdi), %xmm0
+; SSE-NEXT:    movaps 1056(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1200(%rdi), %xmm0
+; SSE-NEXT:    movaps 1152(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1296(%rdi), %xmm0
+; SSE-NEXT:    movaps 1248(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1392(%rdi), %xmm0
+; SSE-NEXT:    movaps 1344(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1488(%rdi), %xmm0
+; SSE-NEXT:    movaps 1440(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1584(%rdi), %xmm0
+; SSE-NEXT:    movaps 1536(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1680(%rdi), %xmm0
+; SSE-NEXT:    movaps 1632(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1776(%rdi), %xmm0
+; SSE-NEXT:    movaps 1728(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1872(%rdi), %xmm0
+; SSE-NEXT:    movaps 1824(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1968(%rdi), %xmm0
+; SSE-NEXT:    movaps 1920(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2064(%rdi), %xmm0
+; SSE-NEXT:    movaps 2016(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2160(%rdi), %xmm0
+; SSE-NEXT:    movaps 2112(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2256(%rdi), %xmm0
+; SSE-NEXT:    movaps 2208(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2352(%rdi), %xmm0
+; SSE-NEXT:    movaps 2304(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2448(%rdi), %xmm0
+; SSE-NEXT:    movaps 2400(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2544(%rdi), %xmm0
+; SSE-NEXT:    movaps 2496(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2640(%rdi), %xmm0
+; SSE-NEXT:    movaps 2592(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2736(%rdi), %xmm0
+; SSE-NEXT:    movaps 2688(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2832(%rdi), %xmm0
+; SSE-NEXT:    movaps 2784(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2928(%rdi), %xmm0
+; SSE-NEXT:    movaps 2880(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3024(%rdi), %xmm0
+; SSE-NEXT:    movaps 2976(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps (%rdi), %xmm1
+; SSE-NEXT:    movaps 48(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 64(%rdi), %xmm0
+; SSE-NEXT:    movaps 16(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 160(%rdi), %xmm0
+; SSE-NEXT:    movaps 112(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 256(%rdi), %xmm0
+; SSE-NEXT:    movaps 208(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 352(%rdi), %xmm0
+; SSE-NEXT:    movaps 304(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 448(%rdi), %xmm0
+; SSE-NEXT:    movaps 400(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 544(%rdi), %xmm0
+; SSE-NEXT:    movaps 496(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 640(%rdi), %xmm0
+; SSE-NEXT:    movaps 592(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 736(%rdi), %xmm0
+; SSE-NEXT:    movaps 688(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 832(%rdi), %xmm0
+; SSE-NEXT:    movaps 784(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 928(%rdi), %xmm0
+; SSE-NEXT:    movaps 880(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1024(%rdi), %xmm0
+; SSE-NEXT:    movaps 976(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1120(%rdi), %xmm0
+; SSE-NEXT:    movaps 1072(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1216(%rdi), %xmm0
+; SSE-NEXT:    movaps 1168(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1312(%rdi), %xmm0
+; SSE-NEXT:    movaps 1264(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1408(%rdi), %xmm0
+; SSE-NEXT:    movaps 1360(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1504(%rdi), %xmm0
+; SSE-NEXT:    movaps 1456(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1600(%rdi), %xmm0
+; SSE-NEXT:    movaps 1552(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1696(%rdi), %xmm0
+; SSE-NEXT:    movaps 1648(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1792(%rdi), %xmm0
+; SSE-NEXT:    movaps 1744(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1888(%rdi), %xmm0
+; SSE-NEXT:    movaps 1840(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1984(%rdi), %xmm0
+; SSE-NEXT:    movaps 1936(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2080(%rdi), %xmm0
+; SSE-NEXT:    movaps 2032(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2176(%rdi), %xmm0
+; SSE-NEXT:    movaps 2128(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2272(%rdi), %xmm0
+; SSE-NEXT:    movaps 2224(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2368(%rdi), %xmm0
+; SSE-NEXT:    movaps 2320(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2464(%rdi), %xmm0
+; SSE-NEXT:    movaps 2416(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2560(%rdi), %xmm0
+; SSE-NEXT:    movaps 2512(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2656(%rdi), %xmm0
+; SSE-NEXT:    movaps 2608(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2752(%rdi), %xmm0
+; SSE-NEXT:    movaps 2704(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2848(%rdi), %xmm0
+; SSE-NEXT:    movaps 2800(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2944(%rdi), %xmm0
+; SSE-NEXT:    movaps 2896(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3040(%rdi), %xmm0
+; SSE-NEXT:    movaps 2992(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%rdi), %xmm0
+; SSE-NEXT:    movaps 32(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 176(%rdi), %xmm0
+; SSE-NEXT:    movaps 128(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 272(%rdi), %xmm0
+; SSE-NEXT:    movaps 224(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 368(%rdi), %xmm0
+; SSE-NEXT:    movaps 320(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 464(%rdi), %xmm0
+; SSE-NEXT:    movaps 416(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 560(%rdi), %xmm0
+; SSE-NEXT:    movaps 512(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 656(%rdi), %xmm0
+; SSE-NEXT:    movaps 608(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 752(%rdi), %xmm0
+; SSE-NEXT:    movaps 704(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 848(%rdi), %xmm0
+; SSE-NEXT:    movaps 800(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 944(%rdi), %xmm0
+; SSE-NEXT:    movaps 896(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1040(%rdi), %xmm0
+; SSE-NEXT:    movaps 992(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1136(%rdi), %xmm0
+; SSE-NEXT:    movaps 1088(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1232(%rdi), %xmm0
+; SSE-NEXT:    movaps 1184(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1328(%rdi), %xmm0
+; SSE-NEXT:    movaps 1280(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1424(%rdi), %xmm0
+; SSE-NEXT:    movaps 1376(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1520(%rdi), %xmm0
+; SSE-NEXT:    movaps 1472(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1616(%rdi), %xmm0
+; SSE-NEXT:    movaps 1568(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1712(%rdi), %xmm0
+; SSE-NEXT:    movaps 1664(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1808(%rdi), %xmm0
+; SSE-NEXT:    movaps 1760(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1904(%rdi), %xmm0
+; SSE-NEXT:    movaps 1856(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2000(%rdi), %xmm0
+; SSE-NEXT:    movaps 1952(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2096(%rdi), %xmm0
+; SSE-NEXT:    movaps 2048(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2192(%rdi), %xmm0
+; SSE-NEXT:    movaps 2144(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2288(%rdi), %xmm0
+; SSE-NEXT:    movaps 2240(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2384(%rdi), %xmm0
+; SSE-NEXT:    movaps 2336(%rdi), %xmm14
+; SSE-NEXT:    movaps %xmm14, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm0[1]
+; SSE-NEXT:    movaps 2480(%rdi), %xmm0
+; SSE-NEXT:    movaps 2432(%rdi), %xmm13
+; SSE-NEXT:    movaps %xmm13, %xmm15
+; SSE-NEXT:    movlhps {{.*#+}} xmm15 = xmm15[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm0[1]
+; SSE-NEXT:    movaps 2576(%rdi), %xmm0
+; SSE-NEXT:    movaps 2528(%rdi), %xmm11
+; SSE-NEXT:    movaps %xmm11, %xmm12
+; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm0[1]
+; SSE-NEXT:    movaps 2672(%rdi), %xmm0
+; SSE-NEXT:    movaps 2624(%rdi), %xmm9
+; SSE-NEXT:    movaps %xmm9, %xmm10
+; SSE-NEXT:    movlhps {{.*#+}} xmm10 = xmm10[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm0[1]
+; SSE-NEXT:    movaps 2768(%rdi), %xmm0
+; SSE-NEXT:    movaps 2720(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm8
+; SSE-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm0[1]
+; SSE-NEXT:    movaps 2864(%rdi), %xmm0
+; SSE-NEXT:    movaps 2816(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm0[1]
+; SSE-NEXT:    movaps 2960(%rdi), %xmm0
+; SSE-NEXT:    movaps 2912(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm0[1]
+; SSE-NEXT:    movaps 3056(%rdi), %xmm0
+; SSE-NEXT:    movaps 3008(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 496(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 480(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 496(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 480(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 496(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 480(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 496(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 480(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r8)
+; SSE-NEXT:    movaps %xmm1, 496(%r9)
+; SSE-NEXT:    movaps %xmm3, 480(%r9)
+; SSE-NEXT:    movaps %xmm6, 464(%r9)
+; SSE-NEXT:    movaps %xmm8, 448(%r9)
+; SSE-NEXT:    movaps %xmm10, 432(%r9)
+; SSE-NEXT:    movaps %xmm12, 416(%r9)
+; SSE-NEXT:    movaps %xmm15, 400(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%r9)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r9)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm2, 496(%rax)
+; SSE-NEXT:    movaps %xmm4, 480(%rax)
+; SSE-NEXT:    movaps %xmm5, 464(%rax)
+; SSE-NEXT:    movaps %xmm7, 448(%rax)
+; SSE-NEXT:    movaps %xmm9, 432(%rax)
+; SSE-NEXT:    movaps %xmm11, 416(%rax)
+; SSE-NEXT:    movaps %xmm13, 400(%rax)
+; SSE-NEXT:    movaps %xmm14, 384(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    addq $2712, %rsp # imm = 0xA98
+; SSE-NEXT:    retq
+;
+; AVX1-ONLY-LABEL: load_i64_stride6_vf64:
+; AVX1-ONLY:       # %bb.0:
+; AVX1-ONLY-NEXT:    subq $3768, %rsp # imm = 0xEB8
+; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 240(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm4[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 672(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovaps 624(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1056(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovaps 1008(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1472(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1440(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1392(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1856(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1824(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1776(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1728(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2240(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2208(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 2160(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2112(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2624(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2592(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 2544(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2496(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3008(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2976(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 2928(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2880(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 432(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 864(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 816(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1280(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1248(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1200(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1632(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1584(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm13[0],xmm14[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2048(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2016(%rdi), %ymm0, %ymm11
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
+; AVX1-ONLY-NEXT:    vmovaps 1968(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps 1920(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm10[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2432(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2400(%rdi), %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX1-ONLY-NEXT:    vmovaps 2352(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 2304(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm5[0],xmm6[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2816(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2784(%rdi), %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vmovaps 2736(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 2688(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm13 = xmm13[1],xmm14[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm9 = xmm9[1],xmm10[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm5[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 736(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 592(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1120(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1024(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 976(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1504(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1360(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1824(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1888(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1792(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1744(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2208(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2272(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 2176(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2128(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2592(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2656(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 2560(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2512(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2976(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 3040(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 2944(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2896(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2784(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2848(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 2752(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2704(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2400(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2464(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 2368(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2320(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2016(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2080(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1984(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1936(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1632(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1696(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1600(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1552(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1312(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vmovaps 1168(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm11[0],xmm12[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 928(%rdi), %ymm0, %ymm9
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm9[0],ymm10[2],ymm9[2]
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 784(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm7[0],xmm8[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 544(%rdi), %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm6[0],ymm15[0],ymm6[2],ymm15[2]
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps 400(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm4[0],xmm5[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm14[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm13[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm14[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm6[1],ymm15[1],ymm6[3],ymm15[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm9[1],ymm10[3],ymm9[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm8[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm11[1],xmm12[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%rdi), %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 320(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 272(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 512(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 464(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 704(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 656(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 896(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 848(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1088(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1040(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1312(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1280(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1232(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1504(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1472(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1424(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1664(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1616(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1888(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1856(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1808(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1760(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2080(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2048(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 2000(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1952(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2272(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2240(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 2192(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2144(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2464(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2432(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 2384(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vmovaps 2336(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm13[0],xmm14[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2656(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2624(%rdi), %ymm0, %ymm11
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
+; AVX1-ONLY-NEXT:    vmovaps 2576(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps 2528(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm10[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2848(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2816(%rdi), %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX1-ONLY-NEXT:    vmovaps 2768(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 2720(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm5[0],xmm6[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3040(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 3008(%rdi), %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vmovaps 2960(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 2912(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd (%rsp), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm13 = xmm13[1],xmm14[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm9 = xmm9[1],xmm10[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm5[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 448(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 384(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 320(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 256(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 192(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 128(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 480(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 416(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 352(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 288(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 224(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 160(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 448(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 384(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 320(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 256(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 192(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 128(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 480(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 416(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 352(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 288(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 224(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 160(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 256(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 320(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 384(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 448(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 480(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 416(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 352(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 288(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 480(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 448(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 416(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 384(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 352(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 320(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 288(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 256(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 224(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 192(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 160(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 128(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 96(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 64(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 480(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 448(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 416(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 384(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 352(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 320(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 288(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 256(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 224(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 192(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 160(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 96(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 64(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%r9)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 480(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm9, 416(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm13, 384(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm15, 352(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 320(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%rax)
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%rax)
+; AVX1-ONLY-NEXT:    addq $3768, %rsp # imm = 0xEB8
+; AVX1-ONLY-NEXT:    vzeroupper
+; AVX1-ONLY-NEXT:    retq
+;
+; AVX2-ONLY-LABEL: load_i64_stride6_vf64:
+; AVX2-ONLY:       # %bb.0:
+; AVX2-ONLY-NEXT:    subq $3416, %rsp # imm = 0xD58
+; AVX2-ONLY-NEXT:    vmovaps 1088(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1056(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 704(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 672(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 240(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm7[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm6[0],ymm1[0],ymm6[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 624(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 576(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm5[0],ymm3[0],ymm5[2],ymm3[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1008(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 960(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1472(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 1392(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1856(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1824(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 1776(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1728(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2240(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2208(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 2160(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2112(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2624(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2592(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 2544(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2496(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3008(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2976(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 2928(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2880(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 48(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 512(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 432(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %xmm13
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm13[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 896(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 864(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 816(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 768(%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm11[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1280(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 1200(%rdi), %xmm14
+; AVX2-ONLY-NEXT:    vmovaps 1152(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm14[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1664(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1632(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 1584(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm7[0],xmm12[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2048(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2016(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 1968(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vmovaps 1920(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm5[0],xmm10[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2432(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovaps 2400(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm8[0],ymm0[2],ymm8[2]
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 2352(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vmovaps 2304(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm6[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2816(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovaps 2784(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 2736(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps 2688(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 296(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 680(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 1064(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 1448(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 1832(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 2216(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 2600(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 2984(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm13[1],mem[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 488(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm11[1],mem[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 872(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm11 = ymm11[1],mem[1],ymm11[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm9[1],xmm14[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 1256(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm9 = ymm9[1],mem[1],ymm9[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm7[1],xmm12[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 1640(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm5[1],xmm10[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 2024(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm6[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 2408(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm8[1],ymm3[3],ymm8[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 2792(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 352(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 208(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 736(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 640(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 592(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 1120(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 1024(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 976(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 1504(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1360(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 1888(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 1792(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1744(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 2272(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 2176(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2128(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 2656(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 2560(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2512(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 3040(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 2944(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2896(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 2848(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 2752(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2704(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 2464(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 2368(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2320(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 2080(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 1984(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1936(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 1696(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 1600(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vmovaps 1552(%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm11[0],xmm12[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 1312(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm0[0],ymm10[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovaps 1168(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm8[0],xmm9[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 928(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm0[0],ymm7[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 832(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vmovaps 784(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm5[0],xmm6[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 544(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm4[0],ymm0[0],ymm4[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %xmm15
+; AVX2-ONLY-NEXT:    vmovaps 400(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm3[0],xmm15[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 160(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 16(%rdi), %xmm14
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm14[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm14[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm13[1],ymm2[3],ymm13[3]
+; AVX2-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %ymm14
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm14[1],ymm1[3],ymm14[3]
+; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm15[1]
+; AVX2-ONLY-NEXT:    vmovaps 544(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm4[1],ymm3[1],ymm4[3],ymm3[3]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vmovaps 736(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm5[1],xmm6[1]
+; AVX2-ONLY-NEXT:    vmovaps 928(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm7[1],ymm2[1],ymm7[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, %ymm15
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vmovaps 1120(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm7[1],ymm1[3],ymm7[3]
+; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm8[1],xmm9[1]
+; AVX2-ONLY-NEXT:    vmovaps 1312(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm10[1],ymm5[1],ymm10[3],ymm5[3]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vmovaps 1504(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, %ymm10
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm11[1],xmm12[1]
+; AVX2-ONLY-NEXT:    vmovaps 1696(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, %ymm9
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vmovaps 1888(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vmovaps 2080(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, %ymm11
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vmovaps 2272(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, %ymm6
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vmovaps 2464(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vmovaps 2656(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vmovaps 2848(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT:    vmovaps 3040(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm13[0],ymm0[2],ymm13[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 80(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm12[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm14[0],ymm0[2],ymm14[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 272(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm12[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 464(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 656(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 608(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm15[0],ymm0[2],ymm15[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 848(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 800(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm7[0],ymm0[2],ymm7[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 1040(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 992(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 1232(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm10[0],ymm0[2],ymm10[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 1424(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 1616(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm8[0],ymm0[2],ymm8[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 1808(%rdi), %xmm14
+; AVX2-ONLY-NEXT:    vmovaps 1760(%rdi), %xmm13
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm13[0],xmm14[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm11[0],ymm0[2],ymm11[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 2000(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vmovaps 1952(%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm11[0],xmm12[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[2],ymm6[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 2192(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vmovaps 2144(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm10[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 2384(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vmovaps 2336(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm7[0],xmm8[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 2576(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vmovaps 2528(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm5[0],xmm6[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 2768(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vmovaps 2720(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm4[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT:    vmovaps 2960(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps 2912(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 136(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 328(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 520(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 712(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 904(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 1096(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 1288(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 1480(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 1672(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 1864(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm13 = xmm13[1],xmm14[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 2056(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm11 = xmm11[1],xmm12[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 2248(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm9 = xmm9[1],xmm10[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 2440(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm7[1],xmm8[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 2632(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm5[1],xmm6[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 2824(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm6 = ymm6[1],mem[1],ymm6[3],mem[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm4[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 3016(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 480(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 416(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 480(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 416(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 480(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 416(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 480(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 416(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 480(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 416(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%r9)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, 480(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 448(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 416(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 384(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm9, 352(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm11, 320(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm13, 288(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm15, 256(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, (%rax)
+; AVX2-ONLY-NEXT:    addq $3416, %rsp # imm = 0xD58
+; AVX2-ONLY-NEXT:    vzeroupper
+; AVX2-ONLY-NEXT:    retq
+;
+; AVX512F-LABEL: load_i64_stride6_vf64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    subq $7240, %rsp # imm = 0x1C48
+; AVX512F-NEXT:    vmovdqa64 2048(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqa64 1280(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqa64 1344(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa64 896(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqa64 960(%rdi), %zmm15
+; AVX512F-NEXT:    vmovdqa64 512(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 128(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqa64 192(%rdi), %zmm2
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,6,0,10,0,6,0,10]
+; AVX512F-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm9
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm8, %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm8, %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm9
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm8, %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm9
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm8, %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm9
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [1,7,0,11,1,7,0,11]
+; AVX512F-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm10, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm10, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm10, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm10, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm11 = [10,4,10,4,10,4,10,4]
+; AVX512F-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm11, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm11, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm11, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm11, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm12 = [11,5,11,5,11,5,11,5]
+; AVX512F-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm12, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm12, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm12, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm12, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm13 = [12,0,0,6,12,0,0,6]
+; AVX512F-NEXT:    # zmm13 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm13, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [13,0,1,7,13,0,1,7]
+; AVX512F-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm3, %zmm6
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm3, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm3, %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm3, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2112(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm8, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm10, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm11, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm12, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm13, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm3, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1664(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa64 1728(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm10, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm11, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm12, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm13, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm3, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2432(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqa64 2496(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm8, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2816(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqa64 2880(%rdi), %zmm1
+; AVX512F-NEXT:    vpermi2q %zmm4, %zmm1, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm5
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm10, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm4, %zmm1, %zmm10
+; AVX512F-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm11, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm1, %zmm4, %zmm11
+; AVX512F-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm12, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm1, %zmm4, %zmm12
+; AVX512F-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm5
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm13, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm1, %zmm4, %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm3, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 448(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 384(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm13 = <0,6,12,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm13, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm14 = <1,7,13,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm14, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm15 = <10,0,6,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} ymm16 = <11,1,7,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm16, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm5 = [4,10,4,10,4,10,4,10]
+; AVX512F-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm5, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [5,11,5,11,5,11,5,11]
+; AVX512F-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 (%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 64(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm13, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm14, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm15, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm16, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1216(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 1152(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm13, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm14, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm15, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm16, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 832(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 768(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm14, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm15, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm16, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm1, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1984(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 1920(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm14, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm15, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm16, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm1, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1600(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 1536(%rdi), %zmm29
+; AVX512F-NEXT:    vmovdqa64 %zmm29, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm29, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm14, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm15, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm16, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm29, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm1, %zmm29
+; AVX512F-NEXT:    vmovdqa64 320(%rdi), %zmm27
+; AVX512F-NEXT:    vmovdqa64 256(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm27, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 704(%rdi), %zmm24
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1088(%rdi), %zmm26
+; AVX512F-NEXT:    vmovdqa64 1024(%rdi), %zmm10
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1472(%rdi), %zmm22
+; AVX512F-NEXT:    vmovdqa64 1408(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm22, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1856(%rdi), %zmm23
+; AVX512F-NEXT:    vmovdqa64 1792(%rdi), %zmm11
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm23, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2240(%rdi), %zmm19
+; AVX512F-NEXT:    vmovdqa64 2176(%rdi), %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2624(%rdi), %zmm21
+; AVX512F-NEXT:    vmovdqa64 2560(%rdi), %zmm12
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2368(%rdi), %zmm25
+; AVX512F-NEXT:    vmovdqa64 2304(%rdi), %zmm20
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm28
+; AVX512F-NEXT:    vpermt2q %zmm25, %zmm13, %zmm28
+; AVX512F-NEXT:    vmovdqa64 3008(%rdi), %zmm18
+; AVX512F-NEXT:    vmovdqa64 2944(%rdi), %zmm9
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm5, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2752(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqa64 2688(%rdi), %zmm17
+; AVX512F-NEXT:    vpermi2q %zmm0, %zmm17, %zmm13
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm30
+; AVX512F-NEXT:    vpermt2q %zmm25, %zmm14, %zmm30
+; AVX512F-NEXT:    vpermi2q %zmm0, %zmm17, %zmm14
+; AVX512F-NEXT:    vmovdqa64 %zmm25, %zmm31
+; AVX512F-NEXT:    vpermt2q %zmm20, %zmm15, %zmm31
+; AVX512F-NEXT:    vpermi2q %zmm17, %zmm0, %zmm15
+; AVX512F-NEXT:    vmovdqa64 %zmm25, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm20, %zmm16, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm17, %zmm0, %zmm16
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm25, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm0, %zmm17, %zmm5
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm1, %zmm17
+; AVX512F-NEXT:    vpermt2q %zmm25, %zmm1, %zmm20
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm27, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm22, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm23, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = [0,0,6,12,0,0,6,12]
+; AVX512F-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm2 = [0,1,7,13,0,1,7,13]
+; AVX512F-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [0,10,0,6,0,10,0,6]
+; AVX512F-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,11,1,7,0,11,1,7]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm0, %zmm24
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm27, %zmm1, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm27, %zmm2, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm3, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm0, %zmm27
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm22, %zmm1, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm22, %zmm2, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm3, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm7
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm1, %zmm7
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm2, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm26, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm3, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm0, %zmm26
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm10
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm1, %zmm10
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm2, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm19, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm3, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm0, %zmm19
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm8
+; AVX512F-NEXT:    vpermt2q %zmm23, %zmm1, %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm25
+; AVX512F-NEXT:    vpermt2q %zmm23, %zmm2, %zmm25
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm11, %zmm3, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm11, %zmm0, %zmm23
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm11
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm1, %zmm11
+; AVX512F-NEXT:    vpermi2q %zmm18, %zmm9, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm2, %zmm4
+; AVX512F-NEXT:    vpermi2q %zmm18, %zmm9, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm6
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm3, %zmm6
+; AVX512F-NEXT:    vpermi2q %zmm9, %zmm18, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm18
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm0, %zmm21
+; AVX512F-NEXT:    movb $56, %al
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    movb $-64, %al
+; AVX512F-NEXT:    kmovw %eax, %k2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm28 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm28 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm13 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm13 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm30 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm30 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm14 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm14 {%k2}
+; AVX512F-NEXT:    movb $24, %al
+; AVX512F-NEXT:    kmovw %eax, %k2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm31 {%k2}
+; AVX512F-NEXT:    movb $-32, %al
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm31 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm7 {%k2}
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm7 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm7 {%k2}
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm7 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm15 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512F-NEXT:    vmovdqa64 %zmm25, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm25
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm16 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm0 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm1 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm2 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm4 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm4 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm6, %zmm6 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm6 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm7 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm7 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm8 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm8 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm5, %zmm9, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm3 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm9 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm10, %zmm10 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm10 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm11, %zmm11 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm26, %zmm11 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm12, %zmm12 # 16-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm19, %zmm12 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm29, %zmm19, %zmm19
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm19 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm17, %zmm22, %zmm17
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm17 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm20, %zmm18, %zmm18
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm18 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm13, 448(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm28, 384(%rsi)
+; AVX512F-NEXT:    vmovups (%rsp), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 320(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 256(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 192(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 128(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 64(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, (%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm14, 448(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 256(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 320(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 128(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 192(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, (%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 64(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm30, 384(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm15, 448(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 256(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 320(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 128(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 192(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, (%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 64(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm31, 384(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm16, 448(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm25, 256(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 320(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 128(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 192(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, (%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 64(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm13, 384(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm5, 448(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm8, 256(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm7, 320(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm6, 128(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm4, 192(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm2, (%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm1, 64(%r9)
+; AVX512F-NEXT:    vmovdqa64 %zmm0, 384(%r9)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovdqa64 %zmm18, 384(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm17, 448(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm19, 256(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm12, 320(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm11, 128(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm10, 192(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm9, (%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm3, 64(%rax)
+; AVX512F-NEXT:    addq $7240, %rsp # imm = 0x1C48
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: load_i64_stride6_vf64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    subq $7240, %rsp # imm = 0x1C48
+; AVX512BW-NEXT:    vmovdqa64 2048(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqa64 1280(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqa64 1344(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm5
+; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm15
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm2
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,6,0,10,0,6,0,10]
+; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm9
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm8, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm8, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm9
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm8, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm9
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm8, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm9
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [1,7,0,11,1,7,0,11]
+; AVX512BW-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm10, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm10, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm10, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm10, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm11 = [10,4,10,4,10,4,10,4]
+; AVX512BW-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm11, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm11, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm11, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm11, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm12 = [11,5,11,5,11,5,11,5]
+; AVX512BW-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm12, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm12, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm12, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm12, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm13 = [12,0,0,6,12,0,0,6]
+; AVX512BW-NEXT:    # zmm13 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm13, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [13,0,1,7,13,0,1,7]
+; AVX512BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm3, %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm3, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm3, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm3, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2112(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm8, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm10, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm11, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm12, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm13, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm3, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1664(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa64 1728(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm8, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm10, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm11, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm12, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm13, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2432(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqa64 2496(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm8, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2816(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqa64 2880(%rdi), %zmm1
+; AVX512BW-NEXT:    vpermi2q %zmm4, %zmm1, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm5
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm10, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm4, %zmm1, %zmm10
+; AVX512BW-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm11, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm4, %zmm11
+; AVX512BW-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm12, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm4, %zmm12
+; AVX512BW-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm13, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm4, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm3, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm13 = <0,6,12,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm13, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm14 = <1,7,13,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm14, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm15 = <10,0,6,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} ymm16 = <11,1,7,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm16, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm5 = [4,10,4,10,4,10,4,10]
+; AVX512BW-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm5, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [5,11,5,11,5,11,5,11]
+; AVX512BW-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm13, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm14, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm16, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm5, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm13, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm14, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm16, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm5, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm14, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm15, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm16, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm1, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1984(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 1920(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm14, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm15, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm16, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm1, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1600(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 1536(%rdi), %zmm29
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm14, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm15, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm16, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm1, %zmm29
+; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm27
+; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm24
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %zmm26
+; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %zmm10
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1472(%rdi), %zmm22
+; AVX512BW-NEXT:    vmovdqa64 1408(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1856(%rdi), %zmm23
+; AVX512BW-NEXT:    vmovdqa64 1792(%rdi), %zmm11
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm23, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2240(%rdi), %zmm19
+; AVX512BW-NEXT:    vmovdqa64 2176(%rdi), %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2624(%rdi), %zmm21
+; AVX512BW-NEXT:    vmovdqa64 2560(%rdi), %zmm12
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2368(%rdi), %zmm25
+; AVX512BW-NEXT:    vmovdqa64 2304(%rdi), %zmm20
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm28
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm13, %zmm28
+; AVX512BW-NEXT:    vmovdqa64 3008(%rdi), %zmm18
+; AVX512BW-NEXT:    vmovdqa64 2944(%rdi), %zmm9
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm18, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2752(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqa64 2688(%rdi), %zmm17
+; AVX512BW-NEXT:    vpermi2q %zmm0, %zmm17, %zmm13
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm30
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm14, %zmm30
+; AVX512BW-NEXT:    vpermi2q %zmm0, %zmm17, %zmm14
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm31
+; AVX512BW-NEXT:    vpermt2q %zmm20, %zmm15, %zmm31
+; AVX512BW-NEXT:    vpermi2q %zmm17, %zmm0, %zmm15
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm20, %zmm16, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm17, %zmm0, %zmm16
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm5, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm0, %zmm17, %zmm5
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm1, %zmm17
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm1, %zmm20
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm23, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm18, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = [0,0,6,12,0,0,6,12]
+; AVX512BW-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm2 = [0,1,7,13,0,1,7,13]
+; AVX512BW-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [0,10,0,6,0,10,0,6]
+; AVX512BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,11,1,7,0,11,1,7]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm0, %zmm24
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm1, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm2, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm3, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm0, %zmm27
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm1, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm2, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm3, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm0, %zmm22
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm7
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm1, %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm2, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm3, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm0, %zmm26
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm10
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm1, %zmm10
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm2, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm3, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm0, %zmm19
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm8
+; AVX512BW-NEXT:    vpermt2q %zmm23, %zmm1, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm25
+; AVX512BW-NEXT:    vpermt2q %zmm23, %zmm2, %zmm25
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm3, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm0, %zmm23
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm11
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm1, %zmm11
+; AVX512BW-NEXT:    vpermi2q %zmm18, %zmm9, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm2, %zmm4
+; AVX512BW-NEXT:    vpermi2q %zmm18, %zmm9, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm6
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm3, %zmm6
+; AVX512BW-NEXT:    vpermi2q %zmm9, %zmm18, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm0, %zmm18
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm0, %zmm21
+; AVX512BW-NEXT:    movb $56, %al
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    movb $-64, %al
+; AVX512BW-NEXT:    kmovd %eax, %k2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm28 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm28 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm13 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm13 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm30 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm30 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm14 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm14 {%k2}
+; AVX512BW-NEXT:    movb $24, %al
+; AVX512BW-NEXT:    kmovd %eax, %k2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm31 {%k2}
+; AVX512BW-NEXT:    movb $-32, %al
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm31 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm7 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm7 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm7 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm7 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm15 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm25
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm16 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm0 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm1 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm2 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm4 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm4 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm6, %zmm6 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm6 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm7 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm7 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm8 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm8 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm5, %zmm9, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm3 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm9 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm10, %zmm10 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm10 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm11, %zmm11 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm11 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm12, %zmm12 # 16-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm12 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm29, %zmm19, %zmm19
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm19 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm17, %zmm22, %zmm17
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm17 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm20, %zmm18, %zmm18
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm18 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, 448(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, 384(%rsi)
+; AVX512BW-NEXT:    vmovups (%rsp), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 320(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 256(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 192(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 128(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 64(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, (%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, 448(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 256(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 320(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 128(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 192(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, (%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 64(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, 384(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, 448(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 256(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 320(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 128(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 192(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, (%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 64(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, 384(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, 448(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, 256(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 320(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 128(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 192(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, (%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 64(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm13, 384(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, 448(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, 256(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, 320(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, 128(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, 192(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, (%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, 64(%r9)
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, 384(%r9)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, 384(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, 448(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, 256(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, 320(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, 128(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, 192(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, (%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, 64(%rax)
+; AVX512BW-NEXT:    addq $7240, %rsp # imm = 0x1C48
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %wide.vec = load <384 x i64>, ptr %in.vec, align 64
+  %strided.vec0 = shufflevector <384 x i64> %wide.vec, <384 x i64> poison, <64 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90, i32 96, i32 102, i32 108, i32 114, i32 120, i32 126, i32 132, i32 138, i32 144, i32 150, i32 156, i32 162, i32 168, i32 174, i32 180, i32 186, i32 192, i32 198, i32 204, i32 210, i32 216, i32 222, i32 228, i32 234, i32 240, i32 246, i32 252, i32 258, i32 264, i32 270, i32 276, i32 282, i32 288, i32 294, i32 300, i32 306, i32 312, i32 318, i32 324, i32 330, i32 336, i32 342, i32 348, i32 354, i32 360, i32 366, i32 372, i32 378>
+  %strided.vec1 = shufflevector <384 x i64> %wide.vec, <384 x i64> poison, <64 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91, i32 97, i32 103, i32 109, i32 115, i32 121, i32 127, i32 133, i32 139, i32 145, i32 151, i32 157, i32 163, i32 169, i32 175, i32 181, i32 187, i32 193, i32 199, i32 205, i32 211, i32 217, i32 223, i32 229, i32 235, i32 241, i32 247, i32 253, i32 259, i32 265, i32 271, i32 277, i32 283, i32 289, i32 295, i32 301, i32 307, i32 313, i32 319, i32 325, i32 331, i32 337, i32 343, i32 349, i32 355, i32 361, i32 367, i32 373, i32 379>
+  %strided.vec2 = shufflevector <384 x i64> %wide.vec, <384 x i64> poison, <64 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92, i32 98, i32 104, i32 110, i32 116, i32 122, i32 128, i32 134, i32 140, i32 146, i32 152, i32 158, i32 164, i32 170, i32 176, i32 182, i32 188, i32 194, i32 200, i32 206, i32 212, i32 218, i32 224, i32 230, i32 236, i32 242, i32 248, i32 254, i32 260, i32 266, i32 272, i32 278, i32 284, i32 290, i32 296, i32 302, i32 308, i32 314, i32 320, i32 326, i32 332, i32 338, i32 344, i32 350, i32 356, i32 362, i32 368, i32 374, i32 380>
+  %strided.vec3 = shufflevector <384 x i64> %wide.vec, <384 x i64> poison, <64 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93, i32 99, i32 105, i32 111, i32 117, i32 123, i32 129, i32 135, i32 141, i32 147, i32 153, i32 159, i32 165, i32 171, i32 177, i32 183, i32 189, i32 195, i32 201, i32 207, i32 213, i32 219, i32 225, i32 231, i32 237, i32 243, i32 249, i32 255, i32 261, i32 267, i32 273, i32 279, i32 285, i32 291, i32 297, i32 303, i32 309, i32 315, i32 321, i32 327, i32 333, i32 339, i32 345, i32 351, i32 357, i32 363, i32 369, i32 375, i32 381>
+  %strided.vec4 = shufflevector <384 x i64> %wide.vec, <384 x i64> poison, <64 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94, i32 100, i32 106, i32 112, i32 118, i32 124, i32 130, i32 136, i32 142, i32 148, i32 154, i32 160, i32 166, i32 172, i32 178, i32 184, i32 190, i32 196, i32 202, i32 208, i32 214, i32 220, i32 226, i32 232, i32 238, i32 244, i32 250, i32 256, i32 262, i32 268, i32 274, i32 280, i32 286, i32 292, i32 298, i32 304, i32 310, i32 316, i32 322, i32 328, i32 334, i32 340, i32 346, i32 352, i32 358, i32 364, i32 370, i32 376, i32 382>
+  %strided.vec5 = shufflevector <384 x i64> %wide.vec, <384 x i64> poison, <64 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95, i32 101, i32 107, i32 113, i32 119, i32 125, i32 131, i32 137, i32 143, i32 149, i32 155, i32 161, i32 167, i32 173, i32 179, i32 185, i32 191, i32 197, i32 203, i32 209, i32 215, i32 221, i32 227, i32 233, i32 239, i32 245, i32 251, i32 257, i32 263, i32 269, i32 275, i32 281, i32 287, i32 293, i32 299, i32 305, i32 311, i32 317, i32 323, i32 329, i32 335, i32 341, i32 347, i32 353, i32 359, i32 365, i32 371, i32 377, i32 383>
+  store <64 x i64> %strided.vec0, ptr %out.vec0, align 64
+  store <64 x i64> %strided.vec1, ptr %out.vec1, align 64
+  store <64 x i64> %strided.vec2, ptr %out.vec2, align 64
+  store <64 x i64> %strided.vec3, ptr %out.vec3, align 64
+  store <64 x i64> %strided.vec4, ptr %out.vec4, align 64
+  store <64 x i64> %strided.vec5, ptr %out.vec5, align 64
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-7.ll
index 547ef4a..49a2bfa 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-7.ll
@@ -124,7 +124,6 @@ define void @load_i64_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vmovdqa %xmm3, (%rax)
 ; AVX512-NEXT:    retq
   %wide.vec = load <14 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <14 x i64> %wide.vec, <14 x i64> poison, <2 x i32> <i32 0, i32 7>
   %strided.vec1 = shufflevector <14 x i64> %wide.vec, <14 x i64> poison, <2 x i32> <i32 1, i32 8>
   %strided.vec2 = shufflevector <14 x i64> %wide.vec, <14 x i64> poison, <2 x i32> <i32 2, i32 9>
@@ -132,7 +131,6 @@ define void @load_i64_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <14 x i64> %wide.vec, <14 x i64> poison, <2 x i32> <i32 4, i32 11>
   %strided.vec5 = shufflevector <14 x i64> %wide.vec, <14 x i64> poison, <2 x i32> <i32 5, i32 12>
   %strided.vec6 = shufflevector <14 x i64> %wide.vec, <14 x i64> poison, <2 x i32> <i32 6, i32 13>
-
   store <2 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i64> %strided.vec2, ptr %out.vec2, align 64
@@ -140,7 +138,6 @@ define void @load_i64_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <2 x i64> %strided.vec4, ptr %out.vec4, align 64
   store <2 x i64> %strided.vec5, ptr %out.vec5, align 64
   store <2 x i64> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -507,7 +504,6 @@ define void @load_i64_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <28 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <28 x i64> %wide.vec, <28 x i64> poison, <4 x i32> <i32 0, i32 7, i32 14, i32 21>
   %strided.vec1 = shufflevector <28 x i64> %wide.vec, <28 x i64> poison, <4 x i32> <i32 1, i32 8, i32 15, i32 22>
   %strided.vec2 = shufflevector <28 x i64> %wide.vec, <28 x i64> poison, <4 x i32> <i32 2, i32 9, i32 16, i32 23>
@@ -515,7 +511,6 @@ define void @load_i64_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <28 x i64> %wide.vec, <28 x i64> poison, <4 x i32> <i32 4, i32 11, i32 18, i32 25>
   %strided.vec5 = shufflevector <28 x i64> %wide.vec, <28 x i64> poison, <4 x i32> <i32 5, i32 12, i32 19, i32 26>
   %strided.vec6 = shufflevector <28 x i64> %wide.vec, <28 x i64> poison, <4 x i32> <i32 6, i32 13, i32 20, i32 27>
-
   store <4 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i64> %strided.vec2, ptr %out.vec2, align 64
@@ -523,7 +518,6 @@ define void @load_i64_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <4 x i64> %strided.vec4, ptr %out.vec4, align 64
   store <4 x i64> %strided.vec5, ptr %out.vec5, align 64
   store <4 x i64> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -1077,7 +1071,6 @@ define void @load_i64_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <56 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <56 x i64> %wide.vec, <56 x i64> poison, <8 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49>
   %strided.vec1 = shufflevector <56 x i64> %wide.vec, <56 x i64> poison, <8 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50>
   %strided.vec2 = shufflevector <56 x i64> %wide.vec, <56 x i64> poison, <8 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51>
@@ -1085,7 +1078,6 @@ define void @load_i64_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <56 x i64> %wide.vec, <56 x i64> poison, <8 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53>
   %strided.vec5 = shufflevector <56 x i64> %wide.vec, <56 x i64> poison, <8 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54>
   %strided.vec6 = shufflevector <56 x i64> %wide.vec, <56 x i64> poison, <8 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55>
-
   store <8 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i64> %strided.vec2, ptr %out.vec2, align 64
@@ -1093,7 +1085,6 @@ define void @load_i64_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <8 x i64> %strided.vec4, ptr %out.vec4, align 64
   store <8 x i64> %strided.vec5, ptr %out.vec5, align 64
   store <8 x i64> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -2199,7 +2190,6 @@ define void @load_i64_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <112 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <112 x i64> %wide.vec, <112 x i64> poison, <16 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49, i32 56, i32 63, i32 70, i32 77, i32 84, i32 91, i32 98, i32 105>
   %strided.vec1 = shufflevector <112 x i64> %wide.vec, <112 x i64> poison, <16 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50, i32 57, i32 64, i32 71, i32 78, i32 85, i32 92, i32 99, i32 106>
   %strided.vec2 = shufflevector <112 x i64> %wide.vec, <112 x i64> poison, <16 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51, i32 58, i32 65, i32 72, i32 79, i32 86, i32 93, i32 100, i32 107>
@@ -2207,7 +2197,6 @@ define void @load_i64_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   %strided.vec4 = shufflevector <112 x i64> %wide.vec, <112 x i64> poison, <16 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53, i32 60, i32 67, i32 74, i32 81, i32 88, i32 95, i32 102, i32 109>
   %strided.vec5 = shufflevector <112 x i64> %wide.vec, <112 x i64> poison, <16 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54, i32 61, i32 68, i32 75, i32 82, i32 89, i32 96, i32 103, i32 110>
   %strided.vec6 = shufflevector <112 x i64> %wide.vec, <112 x i64> poison, <16 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55, i32 62, i32 69, i32 76, i32 83, i32 90, i32 97, i32 104, i32 111>
-
   store <16 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i64> %strided.vec2, ptr %out.vec2, align 64
@@ -2215,7 +2204,6 @@ define void @load_i64_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   store <16 x i64> %strided.vec4, ptr %out.vec4, align 64
   store <16 x i64> %strided.vec5, ptr %out.vec5, align 64
   store <16 x i64> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -4663,7 +4651,6 @@ define void @load_i64_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <224 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <224 x i64> %wide.vec, <224 x i64> poison, <32 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49, i32 56, i32 63, i32 70, i32 77, i32 84, i32 91, i32 98, i32 105, i32 112, i32 119, i32 126, i32 133, i32 140, i32 147, i32 154, i32 161, i32 168, i32 175, i32 182, i32 189, i32 196, i32 203, i32 210, i32 217>
   %strided.vec1 = shufflevector <224 x i64> %wide.vec, <224 x i64> poison, <32 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50, i32 57, i32 64, i32 71, i32 78, i32 85, i32 92, i32 99, i32 106, i32 113, i32 120, i32 127, i32 134, i32 141, i32 148, i32 155, i32 162, i32 169, i32 176, i32 183, i32 190, i32 197, i32 204, i32 211, i32 218>
   %strided.vec2 = shufflevector <224 x i64> %wide.vec, <224 x i64> poison, <32 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51, i32 58, i32 65, i32 72, i32 79, i32 86, i32 93, i32 100, i32 107, i32 114, i32 121, i32 128, i32 135, i32 142, i32 149, i32 156, i32 163, i32 170, i32 177, i32 184, i32 191, i32 198, i32 205, i32 212, i32 219>
@@ -4671,7 +4658,6 @@ define void @load_i64_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   %strided.vec4 = shufflevector <224 x i64> %wide.vec, <224 x i64> poison, <32 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53, i32 60, i32 67, i32 74, i32 81, i32 88, i32 95, i32 102, i32 109, i32 116, i32 123, i32 130, i32 137, i32 144, i32 151, i32 158, i32 165, i32 172, i32 179, i32 186, i32 193, i32 200, i32 207, i32 214, i32 221>
   %strided.vec5 = shufflevector <224 x i64> %wide.vec, <224 x i64> poison, <32 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54, i32 61, i32 68, i32 75, i32 82, i32 89, i32 96, i32 103, i32 110, i32 117, i32 124, i32 131, i32 138, i32 145, i32 152, i32 159, i32 166, i32 173, i32 180, i32 187, i32 194, i32 201, i32 208, i32 215, i32 222>
   %strided.vec6 = shufflevector <224 x i64> %wide.vec, <224 x i64> poison, <32 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55, i32 62, i32 69, i32 76, i32 83, i32 90, i32 97, i32 104, i32 111, i32 118, i32 125, i32 132, i32 139, i32 146, i32 153, i32 160, i32 167, i32 174, i32 181, i32 188, i32 195, i32 202, i32 209, i32 216, i32 223>
-
   store <32 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i64> %strided.vec2, ptr %out.vec2, align 64
@@ -4679,7 +4665,5405 @@ define void @load_i64_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   store <32 x i64> %strided.vec4, ptr %out.vec4, align 64
   store <32 x i64> %strided.vec5, ptr %out.vec5, align 64
   store <32 x i64> %strided.vec6, ptr %out.vec6, align 64
+  ret void
+}
 
+define void @load_i64_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind {
+; SSE-LABEL: load_i64_stride7_vf64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    subq $3240, %rsp # imm = 0xCA8
+; SSE-NEXT:    movapd 208(%rdi), %xmm0
+; SSE-NEXT:    movapd 96(%rdi), %xmm1
+; SSE-NEXT:    movapd 144(%rdi), %xmm2
+; SSE-NEXT:    movapd 192(%rdi), %xmm3
+; SSE-NEXT:    movapd 80(%rdi), %xmm4
+; SSE-NEXT:    movapd 128(%rdi), %xmm5
+; SSE-NEXT:    movapd 176(%rdi), %xmm8
+; SSE-NEXT:    movapd 64(%rdi), %xmm9
+; SSE-NEXT:    movapd (%rdi), %xmm10
+; SSE-NEXT:    movapd 16(%rdi), %xmm7
+; SSE-NEXT:    movapd 32(%rdi), %xmm6
+; SSE-NEXT:    movapd 48(%rdi), %xmm14
+; SSE-NEXT:    movapd 224(%rdi), %xmm11
+; SSE-NEXT:    movapd 112(%rdi), %xmm12
+; SSE-NEXT:    movapd 160(%rdi), %xmm15
+; SSE-NEXT:    movapd %xmm14, %xmm13
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm10[0],xmm13[1]
+; SSE-NEXT:    movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufpd {{.*#+}} xmm10 = xmm10[1],xmm9[0]
+; SSE-NEXT:    movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm7[0],xmm9[1]
+; SSE-NEXT:    movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufpd {{.*#+}} xmm7 = xmm7[1],xmm4[0]
+; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm6[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufpd {{.*#+}} xmm6 = xmm6[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm14[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd %xmm15, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm12[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufpd {{.*#+}} xmm12 = xmm12[1],xmm8[0]
+; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm5[0],xmm8[1]
+; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufpd {{.*#+}} xmm5 = xmm5[1],xmm3[0]
+; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm0[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm15[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 272(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm11[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 288(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm11 = xmm11[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 240(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 304(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 256(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 320(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 336(%rdi), %xmm2
+; SSE-NEXT:    movapd 384(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 400(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 352(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 416(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 368(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 432(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 448(%rdi), %xmm2
+; SSE-NEXT:    movapd 496(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 512(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 464(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 528(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 480(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 544(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 560(%rdi), %xmm2
+; SSE-NEXT:    movapd 608(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 624(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 576(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 640(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 592(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 656(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 672(%rdi), %xmm2
+; SSE-NEXT:    movapd 720(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 736(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 688(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 752(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 704(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 768(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 784(%rdi), %xmm2
+; SSE-NEXT:    movapd 832(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 848(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 800(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 864(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 816(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 880(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 896(%rdi), %xmm2
+; SSE-NEXT:    movapd 944(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 960(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 912(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 976(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 928(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 992(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1008(%rdi), %xmm2
+; SSE-NEXT:    movapd 1056(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1072(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1024(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1088(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1040(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1104(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1120(%rdi), %xmm2
+; SSE-NEXT:    movapd 1168(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1184(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1136(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1200(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1152(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1216(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1232(%rdi), %xmm2
+; SSE-NEXT:    movapd 1280(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1296(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1248(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1312(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1264(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1328(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1344(%rdi), %xmm2
+; SSE-NEXT:    movapd 1392(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1408(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1360(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1424(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1376(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1440(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1456(%rdi), %xmm2
+; SSE-NEXT:    movapd 1504(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1520(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1472(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1536(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1488(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1552(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1568(%rdi), %xmm2
+; SSE-NEXT:    movapd 1616(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1632(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1584(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1648(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1600(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1664(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1680(%rdi), %xmm2
+; SSE-NEXT:    movapd 1728(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1744(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1696(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1760(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1712(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1776(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1792(%rdi), %xmm2
+; SSE-NEXT:    movapd 1840(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1856(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1808(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1872(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1824(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1888(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1904(%rdi), %xmm2
+; SSE-NEXT:    movapd 1952(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1968(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1920(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1984(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 1936(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2000(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2016(%rdi), %xmm2
+; SSE-NEXT:    movapd 2064(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2080(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2032(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2096(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2048(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2112(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2128(%rdi), %xmm2
+; SSE-NEXT:    movapd 2176(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2192(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2144(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2208(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2160(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2224(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2240(%rdi), %xmm2
+; SSE-NEXT:    movapd 2288(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2304(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2256(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2320(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2272(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2336(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2352(%rdi), %xmm2
+; SSE-NEXT:    movapd 2400(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2416(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2368(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2432(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2384(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2448(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2464(%rdi), %xmm2
+; SSE-NEXT:    movapd 2512(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2528(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2480(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2544(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2496(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2560(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2576(%rdi), %xmm2
+; SSE-NEXT:    movapd 2624(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2640(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2592(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2656(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2608(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2672(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2688(%rdi), %xmm2
+; SSE-NEXT:    movapd 2736(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2752(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2704(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2768(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2720(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2784(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2800(%rdi), %xmm2
+; SSE-NEXT:    movapd 2848(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2864(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movapd 2816(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2880(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2832(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2896(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2912(%rdi), %xmm2
+; SSE-NEXT:    movapd 2960(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2976(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2928(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2992(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 2944(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3008(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3024(%rdi), %xmm1
+; SSE-NEXT:    movapd 3072(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm15
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1]
+; SSE-NEXT:    movapd 3088(%rdi), %xmm3
+; SSE-NEXT:    shufpd {{.*#+}} xmm1 = xmm1[1],xmm3[0]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3040(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3104(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3056(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3120(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3136(%rdi), %xmm12
+; SSE-NEXT:    movapd 3184(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm10
+; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm12[0],xmm10[1]
+; SSE-NEXT:    movapd 3200(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm12 = xmm12[1],xmm1[0]
+; SSE-NEXT:    movapd 3152(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3216(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3168(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3232(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3248(%rdi), %xmm9
+; SSE-NEXT:    movapd 3296(%rdi), %xmm0
+; SSE-NEXT:    movapd %xmm0, %xmm6
+; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm9[0],xmm6[1]
+; SSE-NEXT:    movapd 3312(%rdi), %xmm14
+; SSE-NEXT:    shufpd {{.*#+}} xmm9 = xmm9[1],xmm14[0]
+; SSE-NEXT:    movapd 3264(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm2[0],xmm14[1]
+; SSE-NEXT:    movapd 3328(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3280(%rdi), %xmm2
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3344(%rdi), %xmm1
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3360(%rdi), %xmm5
+; SSE-NEXT:    movapd 3408(%rdi), %xmm2
+; SSE-NEXT:    movapd %xmm2, %xmm7
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm5[0],xmm7[1]
+; SSE-NEXT:    movapd 3424(%rdi), %xmm11
+; SSE-NEXT:    shufpd {{.*#+}} xmm5 = xmm5[1],xmm11[0]
+; SSE-NEXT:    movapd 3376(%rdi), %xmm0
+; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
+; SSE-NEXT:    movapd 3440(%rdi), %xmm3
+; SSE-NEXT:    shufpd {{.*#+}} xmm0 = xmm0[1],xmm3[0]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3392(%rdi), %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3456(%rdi), %xmm0
+; SSE-NEXT:    shufpd {{.*#+}} xmm1 = xmm1[1],xmm0[0]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd 3472(%rdi), %xmm2
+; SSE-NEXT:    movapd 3520(%rdi), %xmm4
+; SSE-NEXT:    movapd %xmm4, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd 3536(%rdi), %xmm8
+; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm8[0]
+; SSE-NEXT:    movapd 3488(%rdi), %xmm13
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm13[0],xmm8[1]
+; SSE-NEXT:    movapd 3552(%rdi), %xmm0
+; SSE-NEXT:    shufpd {{.*#+}} xmm13 = xmm13[1],xmm0[0]
+; SSE-NEXT:    movapd 3504(%rdi), %xmm3
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movapd %xmm3, %xmm0
+; SSE-NEXT:    movapd 3568(%rdi), %xmm3
+; SSE-NEXT:    shufpd {{.*#+}} xmm0 = xmm0[1],xmm3[0]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm4[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm1, 496(%rsi)
+; SSE-NEXT:    movapd %xmm7, 480(%rsi)
+; SSE-NEXT:    movapd %xmm6, 464(%rsi)
+; SSE-NEXT:    movapd %xmm10, 448(%rsi)
+; SSE-NEXT:    movapd %xmm15, 432(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 352(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 336(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 320(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 304(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 288(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 272(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 256(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rsi)
+; SSE-NEXT:    movapd %xmm2, 496(%rdx)
+; SSE-NEXT:    movapd %xmm5, 480(%rdx)
+; SSE-NEXT:    movapd %xmm9, 464(%rdx)
+; SSE-NEXT:    movapd %xmm12, 448(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%rdx)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rdx)
+; SSE-NEXT:    movapd %xmm8, 496(%rcx)
+; SSE-NEXT:    movapd %xmm11, 480(%rcx)
+; SSE-NEXT:    movapd %xmm14, 464(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rcx)
+; SSE-NEXT:    movapd %xmm13, 496(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 480(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 496(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 480(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r9)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 496(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 480(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movapd %xmm3, 496(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 480(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 432(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 416(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 368(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    addq $3240, %rsp # imm = 0xCA8
+; SSE-NEXT:    retq
+;
+; AVX1-ONLY-LABEL: load_i64_stride7_vf64:
+; AVX1-ONLY:       # %bb.0:
+; AVX1-ONLY-NEXT:    subq $4264, %rsp # imm = 0x10A8
+; AVX1-ONLY-NEXT:    vmovapd 1216(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm13
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 384(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps 272(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 832(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vmovaps 720(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm14[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1280(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 1120(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1168(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1664(%rdi), %ymm15
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1728(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm15[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 1568(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1616(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2112(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2176(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 2016(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2064(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2560(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2624(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 2464(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2512(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3008(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 3072(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 2912(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2960(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3456(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 3520(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 3360(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3408(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = mem[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 608(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 496(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1056(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 944(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1504(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1392(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1888(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1952(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1792(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vmovaps 1840(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm11[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2336(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2400(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 2240(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps 2288(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm7[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 2784(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 2848(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm6[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 2688(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vmovapd 2736(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm2 = xmm12[0],xmm2[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 3232(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 3296(%rdi), %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovapd 3136(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovapd 3184(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm4 = xmm2[0],xmm4[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 288(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm5[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 384(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm4 = ymm13[0],ymm5[0],ymm13[3],ymm5[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 736(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm4 = xmm14[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 832(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm8 = ymm8[0],ymm9[0],ymm8[3],ymm9[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 1184(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = mem[8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 1280(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm10 = ymm10[0],ymm4[0],ymm10[3],ymm4[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm8[0,1],ymm10[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 1632(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = mem[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 1728(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm13 = ymm15[0],ymm8[0],ymm15[3],ymm8[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 2080(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = mem[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 2176(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm15 = ymm0[0],ymm10[0],ymm0[3],ymm10[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 2528(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 2624(%rdi), %ymm13
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm14 = ymm0[0],ymm13[0],ymm0[3],ymm13[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 2976(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm14 = mem[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 3072(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm15 = ymm0[0],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 3424(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm14 = mem[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 3520(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm15 = ymm0[0],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 160(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm14 = ymm14[0],ymm0[0],ymm14[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 64(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm15 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 608(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm14 = ymm14[0],ymm0[0],ymm14[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 512(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1056(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm14 = ymm1[0],ymm0[0],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 960(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1504(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm14 = ymm1[0],ymm0[0],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1408(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1952(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm15 = ymm0[0],ymm14[0],ymm0[3],ymm14[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1856(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm11 = xmm11[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 2400(%rdi), %ymm11
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm15 = ymm0[0],ymm11[0],ymm0[3],ymm11[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2304(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm7 = xmm7[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 2848(%rdi), %ymm15
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm7 = ymm6[0],ymm15[0],ymm6[3],ymm15[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2752(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm12[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 3296(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm7 = ymm3[0],ymm0[0],ymm3[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 3200(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm2 = xmm2[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm5[3]
+; AVX1-ONLY-NEXT:    vmovapd 240(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm7 = xmm5[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovapd 688(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm7 = xmm7[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovapd 1136(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm4 = xmm4[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT:    vmovapd 1584(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm4 = xmm4[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2144(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm10[3]
+; AVX1-ONLY-NEXT:    vmovapd 2032(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm4 = xmm4[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2592(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm13[3]
+; AVX1-ONLY-NEXT:    vmovapd 2480(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm4 = xmm4[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3040(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 2928(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3488(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 3376(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3264(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovdqa 3152(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2816(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3]
+; AVX1-ONLY-NEXT:    vmovdqa 2704(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2368(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm11[3]
+; AVX1-ONLY-NEXT:    vmovdqa 2256(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1,2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 1920(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm0[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT:    vmovapd 1808(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = xmm2[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 1472(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm4
+; AVX1-ONLY-NEXT:    vblendpd $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm4[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vmovapd 1360(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = xmm4[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 1024(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vblendpd $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = ymm7[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vmovapd 912(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm9[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm10[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 576(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vblendpd $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = ymm7[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vmovapd 464(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm11[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm12[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 128(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm12
+; AVX1-ONLY-NEXT:    vblendpd $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm12 = ymm12[0,1,2],mem[3]
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm14 = xmm13[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm12 = ymm14[0,1],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm12 = ymm13[1],ymm14[0],ymm13[2],ymm14[2]
+; AVX1-ONLY-NEXT:    vmovdqa 192(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm7 = xmm7[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm12[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 304(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[1],ymm12[0],ymm5[2],ymm12[2]
+; AVX1-ONLY-NEXT:    vmovdqa 416(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 528(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm11[1],ymm13[0],ymm11[2],ymm13[2]
+; AVX1-ONLY-NEXT:    vmovdqa 640(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm7 = xmm10[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 752(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm0[1],ymm7[0],ymm0[2],ymm7[2]
+; AVX1-ONLY-NEXT:    vmovdqa 864(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = mem[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 976(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm9[1],ymm10[0],ymm9[2],ymm10[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1088(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm5 = xmm6[8,9,10,11,12,13,14,15],xmm9[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1200(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm6[0],ymm1[2],ymm6[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1312(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm5 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1424(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm4[1],ymm0[0],ymm4[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1536(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm3 = xmm3[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1648(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm4[0],ymm1[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1760(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1872(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm2[1],ymm3[0],ymm2[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1984(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm8[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 2096(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2208(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 2320(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm15[1],ymm0[0],ymm15[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2432(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 2544(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm0[1],ymm15[0],ymm0[2],ymm15[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2656(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 2768(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2880(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 2992(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 3104(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 3216(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 3328(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 3440(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 3552(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 128(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm1 = mem[0],xmm14[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 352(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 256(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm1 = xmm14[0],xmm12[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 576(%rdi), %ymm11
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm11[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 480(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm1 = xmm12[0],xmm13[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 800(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 704(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm1 = xmm13[0],xmm7[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 1024(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 928(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm1 = xmm9[0],xmm10[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 1248(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 1152(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm1 = xmm10[0],xmm6[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 1472(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 1376(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 1696(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 1600(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm1 = xmm1[0],xmm4[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 1920(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 1824(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm1 = xmm1[0],xmm3[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovaps 2144(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 2048(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 2368(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 2272(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 2592(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm8[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 2496(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm1 = xmm1[0],xmm15[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 2816(%rdi), %ymm15
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm15[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 2720(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovapd 3040(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm5[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 2944(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovaps 3264(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 3168(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovaps 3488(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 3392(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 192(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 96(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 416(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 320(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm14[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 544(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm12[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 640(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm11[0],ymm1[0],ymm11[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 864(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 768(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm13[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 992(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm9[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 1088(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm7[0],ymm1[0],ymm7[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1312(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1216(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm10[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 1440(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 1536(%rdi), %ymm13
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm6[0],ymm13[0],ymm6[3],ymm13[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 1760(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovdqa 1664(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 1888(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 1984(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm4[0],ymm10[0],ymm4[3],ymm10[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 2208(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm12[0],ymm0[3],ymm12[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2112(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 2336(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 2432(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm3[0],ymm7[0],ymm3[3],ymm7[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 2656(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm8[0],ymm9[0],ymm8[3],ymm9[2]
+; AVX1-ONLY-NEXT:    vmovdqa 2560(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 2784(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 2880(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm15[0],ymm4[0],ymm15[3],ymm4[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 3104(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm5[0],ymm6[0],ymm5[3],ymm6[2]
+; AVX1-ONLY-NEXT:    vmovdqa 3008(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovdqa 3232(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 3328(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm0[0],ymm3[0],ymm0[3],ymm3[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 3552(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[3],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovdqa 3456(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm13[3]
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm13 = xmm13[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm13[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = mem[0,1,2,3],xmm14[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm10[3]
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm10 = xmm10[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm10[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT:    vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = mem[0,1,2,3],xmm11[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm7[3]
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm7 = xmm7[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm7[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT:    vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = mem[0,1,2,3],xmm8[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm7[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm4 = xmm4[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm4[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT:    vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = mem[0,1,2,3],xmm5[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm3[3]
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} xmm3 = xmm3[0],mem[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT:    vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = mem[0,1,2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 448(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 384(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 320(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 256(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 128(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 480(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 416(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 352(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 288(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 160(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 448(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 384(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 320(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 256(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 128(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 480(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 416(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 352(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 288(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 160(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 256(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 320(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 384(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 448(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 480(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 416(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 352(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 288(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 480(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 448(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 416(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 384(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 352(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 320(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 288(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 256(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 160(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 128(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 32(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 480(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 448(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 416(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 384(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 352(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 320(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 288(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 256(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 160(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 32(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%r9)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 480(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 416(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 384(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 352(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 320(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 288(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 256(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%rax)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovapd %ymm0, 480(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm3, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm4, 416(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm7, 384(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm8, 352(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm11, 320(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm10, 288(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm13, 256(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm14, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm15, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%rax)
+; AVX1-ONLY-NEXT:    addq $4264, %rsp # imm = 0x10A8
+; AVX1-ONLY-NEXT:    vzeroupper
+; AVX1-ONLY-NEXT:    retq
+;
+; AVX2-ONLY-LABEL: load_i64_stride7_vf64:
+; AVX2-ONLY:       # %bb.0:
+; AVX2-ONLY-NEXT:    subq $3976, %rsp # imm = 0xF88
+; AVX2-ONLY-NEXT:    vmovdqa 1216(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 768(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovdqa 320(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovdqa 384(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 224(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vmovdqa 272(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 832(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 672(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vmovdqa 720(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm7[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1280(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1120(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vmovdqa 1168(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1664(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vmovdqa 1728(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm11[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1568(%rdi), %xmm13
+; AVX2-ONLY-NEXT:    vmovdqa 1616(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2112(%rdi), %ymm12
+; AVX2-ONLY-NEXT:    vmovdqa 2176(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm12[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2016(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2064(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2560(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2624(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 2464(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2512(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3008(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3072(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 2912(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2960(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 3456(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 3520(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 3360(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa 3408(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm0[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 96(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 160(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = mem[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 544(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 608(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 448(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 496(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm9[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 992(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1056(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 896(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 944(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm9[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1440(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1504(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1344(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1392(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm9[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1888(%rdi), %ymm14
+; AVX2-ONLY-NEXT:    vmovdqa 1952(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1792(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1840(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2336(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovdqa 2400(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2240(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2288(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm9[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2784(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2848(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2688(%rdi), %xmm15
+; AVX2-ONLY-NEXT:    vmovdqa 2736(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm15[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 3232(%rdi), %ymm10
+; AVX2-ONLY-NEXT:    vmovdqa 3296(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 3136(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovdqa 3184(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm9[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 288(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm5[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 384(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm2 = ymm6[8,9,10,11,12,13,14,15],ymm5[0,1,2,3,4,5,6,7],ymm6[24,25,26,27,28,29,30,31],ymm5[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 736(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm7[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 832(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm2 = ymm4[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1184(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm8[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1280(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = mem[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1632(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = xmm13[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1728(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm2 = ymm11[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2080(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2176(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm2 = ymm12[8,9,10,11,12,13,14,15],ymm11[0,1,2,3,4,5,6,7],ymm12[24,25,26,27,28,29,30,31],ymm11[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2528(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2624(%rdi), %ymm12
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = mem[8,9,10,11,12,13,14,15],ymm12[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm12[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2976(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 3072(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = mem[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 3424(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 3520(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 160(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 64(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 608(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 512(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1056(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 960(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1504(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 1408(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1952(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = ymm14[8,9,10,11,12,13,14,15],ymm13[0,1,2,3,4,5,6,7],ymm14[24,25,26,27,28,29,30,31],ymm13[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 1856(%rdi), %xmm14
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2400(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = ymm3[8,9,10,11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7],ymm3[24,25,26,27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 2304(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2848(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 2752(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm15 = xmm15[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 3296(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = ymm10[8,9,10,11,12,13,14,15],ymm15[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm15[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 3200(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm9 = xmm9[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq 352(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm5[1],ymm0[3],ymm5[3]
+; AVX2-ONLY-NEXT:    vmovdqa 240(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq 800(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm6[1],ymm0[3],ymm6[3]
+; AVX2-ONLY-NEXT:    vmovdqa 688(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq 1248(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm7[1],ymm0[3],ymm7[3]
+; AVX2-ONLY-NEXT:    vmovdqa 1136(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq 1696(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm8[1],ymm0[3],ymm8[3]
+; AVX2-ONLY-NEXT:    vmovdqa 1584(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq 2144(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm11[1],ymm0[3],ymm11[3]
+; AVX2-ONLY-NEXT:    vmovdqa 2032(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq 2592(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm12[1],ymm0[3],ymm12[3]
+; AVX2-ONLY-NEXT:    vmovdqa 2480(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 3040(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps 2928(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 3488(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps 3376(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq 3264(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm15[1],ymm0[3],ymm15[3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = mem[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq 2816(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = mem[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq 2368(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm4[1],ymm0[3],ymm4[3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = mem[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq 1920(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm13[1],ymm0[3],ymm13[3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = mem[0,1],xmm14[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 1472(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = mem[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 1024(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = mem[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 576(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = mem[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 128(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = mem[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 64(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 192(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm12[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 288(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 416(%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 512(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 640(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 736(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 864(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 960(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 1088(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1184(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 1312(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1408(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 1536(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1632(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 1760(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm1 = mem[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1856(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 1984(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm7 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm7, %ymm0, %ymm7
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm7[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2080(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm7 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 2208(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm9 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm9, %ymm0, %ymm9
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm7 = ymm7[2,3],ymm9[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2304(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm9 = mem[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 2432(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm10 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm10, %ymm0, %ymm10
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm9 = ymm9[2,3],ymm10[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2528(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm9 = mem[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 2656(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm13 = mem[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm13, %ymm0, %ymm13
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm9 = ymm9[2,3],ymm13[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2752(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm13 = mem[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 2880(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm14 = mem[8,9,10,11,12,13,14,15],xmm9[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm13[2,3],ymm14[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2976(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm13 = mem[8,9,10,11,12,13,14,15],ymm13[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm13[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 3104(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm14 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm13[2,3],ymm14[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 3200(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm13 = mem[8,9,10,11,12,13,14,15],ymm13[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm13[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 3328(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm14 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm13[2,3],ymm14[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 3424(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm13 = mem[8,9,10,11,12,13,14,15],ymm13[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm13[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 3552(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm14 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-ONLY-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm13[2,3],ymm14[2,3]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-ONLY-NEXT:    vmovdqa 128(%rdi), %ymm14
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm12 = ymm14[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 32(%rdi), %xmm13
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm13 = xmm13[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm11, %ymm0, %ymm11
+; AVX2-ONLY-NEXT:    vmovdqa 352(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm11 = ymm15[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 256(%rdi), %xmm13
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm12 = xmm13[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX2-ONLY-NEXT:    vmovdqa 576(%rdi), %ymm12
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm12[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 480(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm11 = xmm0[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm11[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
+; AVX2-ONLY-NEXT:    vmovdqa 800(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm11[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 704(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm8 = xmm0[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm8[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-ONLY-NEXT:    vmovdqa 1024(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm0[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 928(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm6 = xmm0[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-ONLY-NEXT:    vmovdqa 1248(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm0[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1152(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm0[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vmovdqa 1472(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1376(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm0[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vmovdqa 1696(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1600(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm0[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vmovdqa 1920(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1824(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm2[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vmovdqa 2144(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2048(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm7, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vmovdqa 2368(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2272(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm10, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vmovdqa 2592(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2496(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, %xmm9, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vmovdqa 2816(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2720(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vmovdqa 3040(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2944(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vmovdqa 3264(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 3168(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm0[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vmovdqa 3488(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm8[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 3392(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm0[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 192(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm3 = ymm14[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],ymm14[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 96(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm4 = mem[8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 416(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm3 = ymm15[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],ymm15[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 320(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm4 = xmm13[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 544(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = mem[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 640(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm4 = ymm12[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm12[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 864(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm3 = ymm11[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 768(%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm4 = mem[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 992(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = mem[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1088(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = mem[8,9,10,11,12,13,14,15],ymm5[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm5[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1312(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 1216(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm4 = mem[8,9,10,11,12,13,14,15],xmm12[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1440(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = mem[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1536(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = mem[8,9,10,11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1760(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 1664(%rdi), %xmm13
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 1888(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} xmm2 = xmm2[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 1984(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = mem[8,9,10,11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2208(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 2112(%rdi), %xmm14
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2336(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vpalignr $8, (%rsp), %xmm2, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2432(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = mem[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2656(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 2560(%rdi), %xmm15
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = mem[8,9,10,11,12,13,14,15],xmm15[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 2784(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 2880(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm0 = ymm9[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 3104(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 3008(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 3232(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqa 3328(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm9 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovdqa 3552(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpalignr {{.*#+}} ymm7 = ymm8[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT:    vmovdqa 3456(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm9 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm9 = mem[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm9[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm9 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm9 = mem[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm10 = ymm9[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm6 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
+; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm7 = xmm7[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm6 = ymm6[1],mem[1],ymm6[3],mem[3]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = mem[0,1],xmm11[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm11 = ymm7[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm5 = ymm6[1],ymm5[1],ymm6[3],ymm5[3]
+; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm6 = xmm6[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm6 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm6 = mem[0,1],xmm12[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
+; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = ymm4[1],mem[1],ymm4[3],mem[3]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm5 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm5 = mem[0,1],xmm13[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm9 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm3 = ymm4[1],ymm3[1],ymm4[3],ymm3[3]
+; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm4 = mem[0,1],xmm14[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm12 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm2 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm2 = ymm2[1],mem[1],ymm2[3],mem[3]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = mem[0,1],xmm15[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = mem[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vpunpckhqdq (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = mem[0,1],xmm8[2,3]
+; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 448(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 384(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 320(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 256(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 192(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 128(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 64(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, (%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 480(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 416(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 352(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 288(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 224(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 160(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 96(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 32(%rsi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 448(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 384(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 320(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 256(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 192(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 128(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 64(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, (%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 480(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 416(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 352(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 288(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 224(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 160(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 96(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 32(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, (%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 64(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 128(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 192(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 256(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 320(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 384(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 448(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 480(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 416(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 352(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 288(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 224(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 160(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 96(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 32(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 480(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 448(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 416(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 384(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 352(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 320(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 288(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 256(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 224(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 192(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 160(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 128(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 96(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 64(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 32(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, (%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 480(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 448(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 416(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 384(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 352(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 320(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 288(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 256(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 224(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 192(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 160(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 128(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 96(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 64(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 32(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, (%r9)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 480(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 448(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 416(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 384(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 352(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 320(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 288(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 256(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 160(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, (%rax)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovdqa %ymm0, 480(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm1, 448(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, 416(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm3, 384(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm4, 352(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm5, 320(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm12, 288(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm13, 256(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm9, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm6, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm11, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovdqa %ymm10, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, (%rax)
+; AVX2-ONLY-NEXT:    addq $3976, %rsp # imm = 0xF88
+; AVX2-ONLY-NEXT:    vzeroupper
+; AVX2-ONLY-NEXT:    retq
+;
+; AVX512F-LABEL: load_i64_stride7_vf64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    subq $6536, %rsp # imm = 0x1988
+; AVX512F-NEXT:    vmovdqa64 3328(%rdi), %zmm18
+; AVX512F-NEXT:    vmovdqu64 %zmm18, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 3264(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 3008(%rdi), %zmm19
+; AVX512F-NEXT:    vmovdqa64 2944(%rdi), %zmm20
+; AVX512F-NEXT:    vmovdqa64 2880(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2816(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2752(%rdi), %zmm8
+; AVX512F-NEXT:    vmovdqa64 2688(%rdi), %zmm15
+; AVX512F-NEXT:    vmovdqa64 2432(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2368(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1984(%rdi), %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1920(%rdi), %zmm10
+; AVX512F-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1536(%rdi), %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1472(%rdi), %zmm11
+; AVX512F-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1088(%rdi), %zmm17
+; AVX512F-NEXT:    vmovdqu64 %zmm17, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1024(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %zmm16
+; AVX512F-NEXT:    vmovdqu64 %zmm16, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %zmm12
+; AVX512F-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 192(%rdi), %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 128(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [14,0,0,7,14,0,0,7]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqa 2704(%rdi), %xmm2
+; AVX512F-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm16, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqa 464(%rdi), %xmm2
+; AVX512F-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqa 16(%rdi), %xmm2
+; AVX512F-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm13, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqa 1360(%rdi), %xmm2
+; AVX512F-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqa 912(%rdi), %xmm2
+; AVX512F-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqa 2256(%rdi), %xmm2
+; AVX512F-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqa 1808(%rdi), %xmm2
+; AVX512F-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm18, %zmm4, %zmm0
+; AVX512F-NEXT:    vmovdqa 3152(%rdi), %xmm1
+; AVX512F-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti128 $1, 2880(%rdi), %ymm0, %ymm1
+; AVX512F-NEXT:    vmovdqa 2816(%rdi), %ymm0
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,11]
+; AVX512F-NEXT:    vmovdqu64 %zmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm9, %zmm2
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [0,7,14,0,0,7,14,0]
+; AVX512F-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm11
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm19, %zmm6
+; AVX512F-NEXT:    vmovdqu64 %zmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqa64 3072(%rdi), %zmm30
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm16 = [4,5,6,13,4,5,6,13]
+; AVX512F-NEXT:    # zmm16 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm16, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm30, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti128 $1, 640(%rdi), %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa 576(%rdi), %ymm1
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX512F-NEXT:    vmovdqa64 512(%rdi), %zmm31
+; AVX512F-NEXT:    vmovdqa64 448(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm31, %zmm9, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm31, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 768(%rdi), %zmm18
+; AVX512F-NEXT:    vmovdqa64 704(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm4, %zmm3
+; AVX512F-NEXT:    vmovdqa64 832(%rdi), %zmm26
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm16, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm26, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti128 $1, 192(%rdi), %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa 128(%rdi), %ymm3
+; AVX512F-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX512F-NEXT:    vmovdqa64 64(%rdi), %zmm28
+; AVX512F-NEXT:    vmovdqa64 (%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm9, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm28, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 320(%rdi), %zmm21
+; AVX512F-NEXT:    vmovdqa64 256(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm4, %zmm3
+; AVX512F-NEXT:    vmovdqa64 384(%rdi), %zmm23
+; AVX512F-NEXT:    vpermt2q %zmm23, %zmm16, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm23, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti128 $1, 1536(%rdi), %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa 1472(%rdi), %ymm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX512F-NEXT:    vmovdqa64 1408(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqa64 1344(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm9, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm27
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 1664(%rdi), %zmm10
+; AVX512F-NEXT:    vmovdqa64 1600(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm4, %zmm5
+; AVX512F-NEXT:    vmovdqa64 1728(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm16, %zmm5
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti128 $1, 1088(%rdi), %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa 1024(%rdi), %ymm5
+; AVX512F-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
+; AVX512F-NEXT:    vmovdqa64 960(%rdi), %zmm22
+; AVX512F-NEXT:    vmovdqa64 896(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm22, %zmm9, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 1216(%rdi), %zmm17
+; AVX512F-NEXT:    vmovdqa64 1152(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm4, %zmm5
+; AVX512F-NEXT:    vmovdqa64 1280(%rdi), %zmm19
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm16, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti128 $1, 2432(%rdi), %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa 2368(%rdi), %ymm5
+; AVX512F-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
+; AVX512F-NEXT:    vmovdqa64 2304(%rdi), %zmm20
+; AVX512F-NEXT:    vmovdqa64 2240(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm20, %zmm9, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm20, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 2560(%rdi), %zmm24
+; AVX512F-NEXT:    vmovdqa64 2496(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm4, %zmm5
+; AVX512F-NEXT:    vmovdqa64 2624(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm16, %zmm5
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti128 $1, 1984(%rdi), %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa 1920(%rdi), %ymm5
+; AVX512F-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
+; AVX512F-NEXT:    vmovdqa64 1856(%rdi), %zmm14
+; AVX512F-NEXT:    vmovdqa64 1792(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm9, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 2112(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqa64 2048(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm4, %zmm7
+; AVX512F-NEXT:    vmovdqa64 2176(%rdi), %zmm13
+; AVX512F-NEXT:    vpermt2q %zmm13, %zmm16, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm5, %zmm7, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti128 $1, 3328(%rdi), %ymm0, %ymm5
+; AVX512F-NEXT:    vmovdqa 3264(%rdi), %ymm7
+; AVX512F-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm5 = ymm7[0,1,2,3,4,5],ymm5[6,7]
+; AVX512F-NEXT:    vmovdqa64 3200(%rdi), %zmm12
+; AVX512F-NEXT:    vmovdqa64 3136(%rdi), %zmm7
+; AVX512F-NEXT:    vpermi2q %zmm12, %zmm7, %zmm9
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm25
+; AVX512F-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm12
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3],ymm5[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 3456(%rdi), %zmm9
+; AVX512F-NEXT:    vmovdqa64 3392(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm9, %zmm7, %zmm4
+; AVX512F-NEXT:    vmovdqa64 3520(%rdi), %zmm29
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm16, %zmm4
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm5, %zmm4, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa 2880(%rdi), %ymm4
+; AVX512F-NEXT:    vpalignr {{.*#+}} ymm4 = ymm0[8,9,10,11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7],ymm0[24,25,26,27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm0 = [5,12]
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm0, %zmm5
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm7 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [9,0,7,0,9,0,7,0]
+; AVX512F-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm11, %zmm4, %zmm6
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [4,5,6,14,4,5,6,14]
+; AVX512F-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm5, %zmm6
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm7, %zmm6, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa 640(%rdi), %ymm7
+; AVX512F-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],ymm1[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm31, %zmm0, %zmm7
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm7
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm4, %zmm7
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm5, %zmm7
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm7, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa 192(%rdi), %ymm1
+; AVX512F-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm0, %zmm6
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm6
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm4, %zmm6
+; AVX512F-NEXT:    vpermt2q %zmm23, %zmm5, %zmm6
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm6, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa 1536(%rdi), %ymm1
+; AVX512F-NEXT:    vpalignr {{.*#+}} ymm1 = ymm3[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm3[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm27, %zmm0, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm4, %zmm3
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm5, %zmm3 # 64-byte Folded Reload
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa 1088(%rdi), %ymm1
+; AVX512F-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm22, %zmm0, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm17, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm23 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm23, %zmm4, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm5, %zmm3
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa 2432(%rdi), %ymm1
+; AVX512F-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm20, %zmm0, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm4, %zmm3
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm5, %zmm3 # 64-byte Folded Reload
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa 1984(%rdi), %ymm1
+; AVX512F-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm0, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm22, %zmm4, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm13, %zmm5, %zmm3
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa 3328(%rdi), %ymm1
+; AVX512F-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vpermi2q %zmm25, %zmm12, %zmm0
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2q %zmm31, %zmm9, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm5, %zmm4
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm4, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm15 = [10,3,10,3,10,3,10,3]
+; AVX512F-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm26
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm15, %zmm26
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm20 = [11,4,11,4,11,4,11,4]
+; AVX512F-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm20, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm25 = [12,5,12,5,12,5,12,5]
+; AVX512F-NEXT:    # zmm25 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm25, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm8 = [13,6,13,6,13,6,13,6]
+; AVX512F-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm8, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm14
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [2,9,2,9,2,9,2,9]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm19
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm12
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm15, %zmm12
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm20, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm25, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm8, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm0, %zmm19
+; AVX512F-NEXT:    vmovdqu64 %zmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm15, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm19
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm20, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm25, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm8, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm0, %zmm28
+; AVX512F-NEXT:    vmovdqu64 %zmm28, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm15, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm20, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm25, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm8, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm0, %zmm23
+; AVX512F-NEXT:    vmovdqu64 %zmm23, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm5
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm15, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm20, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm25, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm8, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm0, %zmm30
+; AVX512F-NEXT:    vmovdqu64 %zmm30, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm15, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm20, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm25, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm8, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm27
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm15, %zmm27
+; AVX512F-NEXT:    vpermi2q %zmm9, %zmm31, %zmm15
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm28
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm20, %zmm28
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm2 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm16, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm8, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm23 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm23, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm9, %zmm31, %zmm20
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm24
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm25, %zmm24
+; AVX512F-NEXT:    vpermi2q %zmm9, %zmm31, %zmm25
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm8, %zmm2
+; AVX512F-NEXT:    vpermi2q %zmm9, %zmm31, %zmm8
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm31
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,5,6,9,0,5,6,9]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm0, %zmm12
+; AVX512F-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm26
+; AVX512F-NEXT:    vmovdqu64 %zmm26, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm0, %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm19
+; AVX512F-NEXT:    vmovdqu64 %zmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm0, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm0, %zmm27
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm0, %zmm15
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,5,6,10,0,5,6,10]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm0, %zmm28
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm0, %zmm20
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,5,6,11,0,5,6,11]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm0, %zmm24
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm0, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm0, %zmm25
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,5,6,12,0,5,6,12]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm0, %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm0, %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm0, %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm0, %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm0, %zmm8
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [4,5,8,15,4,5,8,15]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm0, %zmm31
+; AVX512F-NEXT:    vmovdqu64 %zmm31, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    movb $24, %al
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm9 = <0,7,14,u>
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm9, %zmm19
+; AVX512F-NEXT:    vshufi64x2 {{.*#+}} zmm19 {%k1} = zmm11[0,1,4,5],zmm6[4,5,0,1]
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm2 = [7,0,9,0,7,0,9,0]
+; AVX512F-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm11, %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,u,4,11>
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm3, %zmm11
+; AVX512F-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm29 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm9, %zmm29
+; AVX512F-NEXT:    vshufi64x2 {{.*#+}} zmm29 {%k1} = zmm13[0,1,4,5],zmm18[4,5,0,1]
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm13, %zmm2, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm3, %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm18 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vshufi64x2 {{.*#+}} zmm18 {%k1} = zmm0[0,1,4,5],zmm16[4,5,0,1]
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm2, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm16, %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm9, %zmm26
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vshufi64x2 {{.*#+}} zmm26 {%k1} = zmm0[0,1,4,5],zmm17[4,5,0,1]
+; AVX512F-NEXT:    vmovdqa64 %zmm17, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm2, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm22, %zmm9, %zmm31
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm0
+; AVX512F-NEXT:    vshufi64x2 {{.*#+}} zmm31 {%k1} = zmm7[0,1,4,5],zmm21[4,5,0,1]
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm2, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm9, %zmm17
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vshufi64x2 {{.*#+}} zmm17 {%k1} = zmm0[0,1,4,5],zmm23[4,5,0,1]
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm2, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm23, %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm9, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2q %zmm14, %zmm0, %zmm9
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vshufi64x2 {{.*#+}} zmm9 {%k1} = zmm0[0,1,4,5],zmm30[4,5,0,1]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm23
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm2, %zmm23
+; AVX512F-NEXT:    vpermi2q %zmm0, %zmm30, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vshufi64x2 {{.*#+}} zmm13 {%k1} = zmm1[0,1,4,5],zmm5[4,5,0,1]
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm3, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm0 = <9,0,7,u>
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm16
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm16
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} xmm30 = [6,13]
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm30, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm12
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm12
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm30, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm10
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm10
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm30, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm0, %zmm6
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm30, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm5
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm0, %zmm5
+; AVX512F-NEXT:    vpermt2q %zmm22, %zmm30, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm0, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm30, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm11, %zmm0, %zmm4
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2q %zmm1, %zmm14, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm30, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm30, %zmm11
+; AVX512F-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vinsertf64x4 $0, %ymm1, %zmm7, %zmm1
+; AVX512F-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovups (%rsp), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vinsertf64x4 $0, %ymm1, %zmm7, %zmm1
+; AVX512F-NEXT:    vmovups %zmm1, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vinsertf64x4 $0, %ymm1, %zmm7, %zmm1
+; AVX512F-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vinsertf64x4 $0, %ymm1, %zmm7, %zmm1
+; AVX512F-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    movb $-32, %al
+; AVX512F-NEXT:    kmovw %eax, %k2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm29 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm19 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm26 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm18 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm17 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm31 {%k2}
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm13 {%k2}
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm9 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm4 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm4 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm16 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm16 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm12 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm12 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm10 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm10 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm6 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm6 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm5 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm5 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm3 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm0 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm0 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm1 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm30 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm1 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm1 {%k2}
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm27 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm24 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm20 {%k2}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm25, %zmm7 {%k2}
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm14 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa 2752(%rdi), %ymm1
+; AVX512F-NEXT:    vpalignr {{.*#+}} ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm1, %zmm23, %zmm28
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm28 {%k2}
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vinsertf64x4 $0, %ymm14, %zmm1, %zmm1
+; AVX512F-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa 512(%rdi), %ymm1
+; AVX512F-NEXT:    vpalignr {{.*#+}} ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm1, %zmm14, %zmm21
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm21 {%k2}
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm15 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa 64(%rdi), %ymm1
+; AVX512F-NEXT:    vpalignr {{.*#+}} ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm1, %zmm14, %zmm22
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm22 {%k2}
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vinsertf64x4 $0, %ymm15, %zmm1, %zmm1
+; AVX512F-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa 1408(%rdi), %ymm1
+; AVX512F-NEXT:    vpalignr {{.*#+}} ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm1, %zmm14, %zmm23
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm23 {%k2}
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm15 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa 960(%rdi), %ymm1
+; AVX512F-NEXT:    vpalignr {{.*#+}} ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm1, %zmm14, %zmm25
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm25 {%k2}
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vinsertf64x4 $0, %ymm15, %zmm1, %zmm1
+; AVX512F-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa 2304(%rdi), %ymm15
+; AVX512F-NEXT:    vpalignr {{.*#+}} ymm15 = mem[8,9,10,11,12,13,14,15],ymm15[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm15[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vextracti128 $1, %ymm15, %xmm15
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm15, %zmm1, %zmm15
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm15 {%k2}
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm11 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa 1856(%rdi), %ymm1
+; AVX512F-NEXT:    vpalignr {{.*#+}} ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm1, %zmm14, %zmm1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm14, %zmm1 {%k2}
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vinsertf64x4 $0, %ymm11, %zmm14, %zmm14
+; AVX512F-NEXT:    vmovups %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa 3200(%rdi), %ymm14
+; AVX512F-NEXT:    vpalignr {{.*#+}} ymm14 = mem[8,9,10,11,12,13,14,15],ymm14[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm14[16,17,18,19,20,21,22,23]
+; AVX512F-NEXT:    vextracti128 $1, %ymm14, %xmm14
+; AVX512F-NEXT:    vinserti32x4 $0, %xmm14, %zmm2, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm2 {%k2}
+; AVX512F-NEXT:    vmovdqa64 %zmm9, 448(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm13, 384(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm31, 320(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm17, 256(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm18, 192(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm26, 128(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm19, 64(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm29, (%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm0, 448(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm3, 256(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm5, 320(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm6, 128(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm10, 192(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm12, (%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm16, 64(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm4, 384(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm7, 448(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm20, 256(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm24, 320(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm27, 128(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm30, 64(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 384(%rcx)
+; AVX512F-NEXT:    vmovdqa64 %zmm2, 448(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm1, 256(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm15, 320(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm25, 128(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm23, 192(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm22, (%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm21, 64(%r8)
+; AVX512F-NEXT:    vmovdqa64 %zmm28, 384(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 448(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 256(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 320(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 384(%r9)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 448(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 256(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 320(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 384(%rax)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 384(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 448(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 256(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 320(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%rax)
+; AVX512F-NEXT:    vmovups (%rsp), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%rax)
+; AVX512F-NEXT:    addq $6536, %rsp # imm = 0x1988
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: load_i64_stride7_vf64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    subq $6536, %rsp # imm = 0x1988
+; AVX512BW-NEXT:    vmovdqa64 3328(%rdi), %zmm19
+; AVX512BW-NEXT:    vmovdqu64 %zmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 3264(%rdi), %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 3008(%rdi), %zmm16
+; AVX512BW-NEXT:    vmovdqa64 2944(%rdi), %zmm28
+; AVX512BW-NEXT:    vmovdqa64 2880(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2816(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2752(%rdi), %zmm20
+; AVX512BW-NEXT:    vmovdqa64 2688(%rdi), %zmm11
+; AVX512BW-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2432(%rdi), %zmm17
+; AVX512BW-NEXT:    vmovdqu64 %zmm17, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2368(%rdi), %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1984(%rdi), %zmm18
+; AVX512BW-NEXT:    vmovdqu64 %zmm18, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1920(%rdi), %zmm10
+; AVX512BW-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1536(%rdi), %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1472(%rdi), %zmm12
+; AVX512BW-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm15
+; AVX512BW-NEXT:    vmovdqu64 %zmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [14,0,0,7,14,0,0,7]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa 2704(%rdi), %xmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa 464(%rdi), %xmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa 16(%rdi), %xmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa 1360(%rdi), %xmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa 912(%rdi), %xmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa 2256(%rdi), %xmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm18, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa 1808(%rdi), %xmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm2, %zmm1, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm19, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqa 3152(%rdi), %xmm1
+; AVX512BW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti128 $1, 2880(%rdi), %ymm0, %ymm0
+; AVX512BW-NEXT:    vmovdqa 2816(%rdi), %ymm1
+; AVX512BW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm7 = [4,11]
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm20, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm20, %zmm7, %zmm1
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [0,7,14,0,0,7,14,0]
+; AVX512BW-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm8
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm15, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 3072(%rdi), %zmm30
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm18 = [4,5,6,13,4,5,6,13]
+; AVX512BW-NEXT:    # zmm18 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm18, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm30, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti128 $1, 640(%rdi), %ymm0, %ymm0
+; AVX512BW-NEXT:    vmovdqa 576(%rdi), %ymm1
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm7, %zmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm16
+; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm15, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm27
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm18, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm27, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti128 $1, 192(%rdi), %ymm0, %ymm0
+; AVX512BW-NEXT:    vmovdqa 128(%rdi), %ymm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm26
+; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm7, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm26, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm22
+; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm15, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm25
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm18, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm25, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti128 $1, 1536(%rdi), %ymm0, %ymm0
+; AVX512BW-NEXT:    vmovdqa 1472(%rdi), %ymm3
+; AVX512BW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX512BW-NEXT:    vmovdqa64 1408(%rdi), %zmm23
+; AVX512BW-NEXT:    vmovdqa64 1344(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm23, %zmm7, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm23, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 1664(%rdi), %zmm21
+; AVX512BW-NEXT:    vmovdqa64 1600(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm15, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 1728(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm18, %zmm4
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm4, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti128 $1, 1088(%rdi), %ymm0, %ymm0
+; AVX512BW-NEXT:    vmovdqa 1024(%rdi), %ymm3
+; AVX512BW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm20
+; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm20, %zmm7, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm20, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %zmm12
+; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm15, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 1280(%rdi), %zmm19
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm18, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm4, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti128 $1, 2432(%rdi), %ymm0, %ymm0
+; AVX512BW-NEXT:    vmovdqa 2368(%rdi), %ymm3
+; AVX512BW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX512BW-NEXT:    vmovdqa64 2304(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 2240(%rdi), %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm7, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm31
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 2560(%rdi), %zmm17
+; AVX512BW-NEXT:    vmovdqa64 2496(%rdi), %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm15, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 2624(%rdi), %zmm9
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm18, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm5, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti128 $1, 1984(%rdi), %ymm0, %ymm0
+; AVX512BW-NEXT:    vmovdqa 1920(%rdi), %ymm5
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX512BW-NEXT:    vmovdqa64 1856(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqa64 1792(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm7, %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 2112(%rdi), %zmm24
+; AVX512BW-NEXT:    vmovdqa64 2048(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm15, %zmm6
+; AVX512BW-NEXT:    vmovdqa64 2176(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm18, %zmm6
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm6, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti128 $1, 3328(%rdi), %ymm0, %ymm6
+; AVX512BW-NEXT:    vmovdqa 3264(%rdi), %ymm0
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm6[6,7]
+; AVX512BW-NEXT:    vmovdqa64 3200(%rdi), %zmm14
+; AVX512BW-NEXT:    vmovdqa64 3136(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm14, %zmm3, %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 3456(%rdi), %zmm14
+; AVX512BW-NEXT:    vmovdqa64 3392(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm14, %zmm7, %zmm15
+; AVX512BW-NEXT:    vmovdqa64 3520(%rdi), %zmm29
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm18, %zmm15
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm6, %zmm15, %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa 2880(%rdi), %ymm6
+; AVX512BW-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm6 = mem[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm7 = [5,12]
+; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm7, %zmm11
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm6[4,5,6,7]
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [9,0,7,0,9,0,7,0]
+; AVX512BW-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm11
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm15
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm6, %zmm11
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm8
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm13 = [4,5,6,14,4,5,6,14]
+; AVX512BW-NEXT:    # zmm13 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm13, %zmm11
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm10, %zmm11, %zmm10
+; AVX512BW-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa 640(%rdi), %ymm10
+; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[8,9,10,11,12,13,14,15],ymm10[0,1,2,3,4,5,6,7],ymm1[24,25,26,27,28,29,30,31],ymm10[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm10 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm10
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm6, %zmm10
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm13, %zmm10
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm10, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa 192(%rdi), %ymm1
+; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm2[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm2[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm7, %zmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm6, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm13, %zmm2
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa 1536(%rdi), %ymm1
+; AVX512BW-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm23, %zmm7, %zmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm6, %zmm2
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm13, %zmm2 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa 1088(%rdi), %ymm1
+; AVX512BW-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm20, %zmm7, %zmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm6, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm13, %zmm2
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa 2432(%rdi), %ymm1
+; AVX512BW-NEXT:    vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm31, %zmm7, %zmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm6, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm13, %zmm2
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa 1984(%rdi), %ymm1
+; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm5[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm7, %zmm2
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm6, %zmm2
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm13, %zmm2 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa 3328(%rdi), %ymm1
+; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm0[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm3, %zmm1, %zmm7
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm4, %zmm14, %zmm6
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm13, %zmm6
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm6, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm18 = [10,3,10,3,10,3,10,3]
+; AVX512BW-NEXT:    # zmm18 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm18, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm19 = [11,4,11,4,11,4,11,4]
+; AVX512BW-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm19, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm20 = [12,5,12,5,12,5,12,5]
+; AVX512BW-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm20, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm9 = [13,6,13,6,13,6,13,6]
+; AVX512BW-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm9, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm3 = [2,9,2,9,2,9,2,9]
+; AVX512BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm11
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm18, %zmm11
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm19, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm20, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm9, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm23
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm18, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm19, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm20, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm9, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm3, %zmm25
+; AVX512BW-NEXT:    vmovdqu64 %zmm25, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm18, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm19, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm20, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm22
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm9, %zmm22
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm3, %zmm26
+; AVX512BW-NEXT:    vmovdqu64 %zmm26, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm18, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm19, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm20, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm9, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm25
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm3, %zmm27
+; AVX512BW-NEXT:    vmovdqu64 %zmm27, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm7
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm18, %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm19, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm20, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm9, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm3, %zmm30
+; AVX512BW-NEXT:    vmovdqu64 %zmm30, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm24
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm31
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm18, %zmm31
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm6
+; AVX512BW-NEXT:    vpermi2q %zmm14, %zmm4, %zmm18
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm19, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm5 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm9, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm9, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm9, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm9, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm5
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm5 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm9, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm14, %zmm6, %zmm19
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm5 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm28
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm20, %zmm28
+; AVX512BW-NEXT:    vpermi2q %zmm14, %zmm6, %zmm20
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm5
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm9, %zmm5
+; AVX512BW-NEXT:    vpermi2q %zmm14, %zmm6, %zmm9
+; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm3, %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm3, %zmm24
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,5,6,9,0,5,6,9]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm0, %zmm11
+; AVX512BW-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm0, %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm0, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm0, %zmm31
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm0, %zmm18
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,5,6,10,0,5,6,10]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm0, %zmm19
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,5,6,11,0,5,6,11]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm0, %zmm28
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm0, %zmm20
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,5,6,12,0,5,6,12]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm0, %zmm23
+; AVX512BW-NEXT:    vmovdqu64 %zmm23, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm0, %zmm22
+; AVX512BW-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm0, %zmm25
+; AVX512BW-NEXT:    vmovdqu64 %zmm25, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm0, %zmm9
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [4,5,8,15,4,5,8,15]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm0, %zmm24
+; AVX512BW-NEXT:    movb $24, %al
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm14 = <0,7,14,u>
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm27 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm14, %zmm27
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm1
+; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm27 {%k1} = zmm8[0,1,4,5],zmm26[4,5,0,1]
+; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [7,0,9,0,7,0,9,0]
+; AVX512BW-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm6, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = <u,u,4,11>
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm29 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm14, %zmm29
+; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm29 {%k1} = zmm10[0,1,4,5],zmm16[4,5,0,1]
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm6, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm2, %zmm10
+; AVX512BW-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm14, %zmm26
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm1
+; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm26 {%k1} = zmm13[0,1,4,5],zmm17[4,5,0,1]
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm6, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm14, %zmm17
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm17 {%k1} = zmm0[0,1,4,5],zmm15[4,5,0,1]
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm6, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm14, %zmm16
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm16 {%k1} = zmm12[0,1,4,5],zmm0[4,5,0,1]
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm6, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm2, %zmm12
+; AVX512BW-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm14, %zmm15
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm15 {%k1} = zmm0[0,1,4,5],zmm21[4,5,0,1]
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm6, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm14, %zmm12
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm21, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm14 {%k1} = zmm0[0,1,4,5],zmm4[4,5,0,1]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm23
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm6, %zmm23
+; AVX512BW-NEXT:    vpermi2q %zmm0, %zmm4, %zmm6
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm12 {%k1} = zmm1[0,1,4,5],zmm5[4,5,0,1]
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = <9,0,7,u>
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm2, %zmm13
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm3 = [6,13]
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm2, %zmm11
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm2, %zmm10
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm2, %zmm8
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm30
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm2, %zmm30
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm2, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm2, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm0, %zmm21, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovups (%rsp), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vinsertf64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovups %zmm1, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vinsertf64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vinsertf64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vinsertf64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vinsertf64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vinsertf64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vinsertf64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm24, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    movb $-32, %al
+; AVX512BW-NEXT:    kmovd %eax, %k2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm29 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm27 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm17 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm26 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm12 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm14 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm5 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm5 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm13 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm13 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm10 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm10 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm30 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm30 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm2 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm31 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm28 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm7 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm0 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm0 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm0 {%k2}
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2752(%rdi), %ymm18
+; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm18 = mem[8,9,10,11,12,13,14,15],ymm18[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm18[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vextracti32x4 $1, %ymm18, %xmm18
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm18, %zmm23, %zmm18
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm18 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %ymm19
+; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm19 = mem[8,9,10,11,12,13,14,15],ymm19[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm19[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vextracti32x4 $1, %ymm19, %xmm19
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm19, %zmm0, %zmm19
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm19 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %ymm20
+; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm20 = mem[8,9,10,11,12,13,14,15],ymm20[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm20[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vextracti32x4 $1, %ymm20, %xmm20
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm20, %zmm0, %zmm20
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 1408(%rdi), %ymm21
+; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm21 = mem[8,9,10,11,12,13,14,15],ymm21[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm21[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vextracti32x4 $1, %ymm21, %xmm21
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm21, %zmm0, %zmm21
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %ymm22
+; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm22 = mem[8,9,10,11,12,13,14,15],ymm22[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm22[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vextracti32x4 $1, %ymm22, %xmm22
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm22, %zmm0, %zmm22
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm22 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 2304(%rdi), %ymm23
+; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm23 = mem[8,9,10,11,12,13,14,15],ymm23[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm23[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vextracti32x4 $1, %ymm23, %xmm23
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm23, %zmm0, %zmm23
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm23 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 1856(%rdi), %ymm24
+; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm24 = mem[8,9,10,11,12,13,14,15],ymm24[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm24[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vextracti32x4 $1, %ymm24, %xmm24
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm24, %zmm0, %zmm24
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm24 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 3200(%rdi), %ymm25
+; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm25 = mem[8,9,10,11,12,13,14,15],ymm25[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm25[16,17,18,19,20,21,22,23]
+; AVX512BW-NEXT:    vextracti32x4 $1, %ymm25, %xmm25
+; AVX512BW-NEXT:    vinserti32x4 $0, %xmm25, %zmm6, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm0 {%k2}
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, 448(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, 384(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, 320(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, 256(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, 192(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, 128(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, 64(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, (%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, 448(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, 256(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, 320(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, 128(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, 192(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, (%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, 64(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, 384(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm2, 448(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm2, 256(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm2, 320(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, 128(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, 192(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, (%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, 64(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, 384(%rcx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, 448(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, 256(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, 320(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, 128(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, 192(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, (%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, 64(%r8)
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, 384(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 448(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 256(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 320(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 384(%r9)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 448(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 256(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 320(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 384(%rax)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 384(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 448(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 256(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 320(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%rax)
+; AVX512BW-NEXT:    vmovups (%rsp), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%rax)
+; AVX512BW-NEXT:    addq $6536, %rsp # imm = 0x1988
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %wide.vec = load <448 x i64>, ptr %in.vec, align 64
+  %strided.vec0 = shufflevector <448 x i64> %wide.vec, <448 x i64> poison, <64 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49, i32 56, i32 63, i32 70, i32 77, i32 84, i32 91, i32 98, i32 105, i32 112, i32 119, i32 126, i32 133, i32 140, i32 147, i32 154, i32 161, i32 168, i32 175, i32 182, i32 189, i32 196, i32 203, i32 210, i32 217, i32 224, i32 231, i32 238, i32 245, i32 252, i32 259, i32 266, i32 273, i32 280, i32 287, i32 294, i32 301, i32 308, i32 315, i32 322, i32 329, i32 336, i32 343, i32 350, i32 357, i32 364, i32 371, i32 378, i32 385, i32 392, i32 399, i32 406, i32 413, i32 420, i32 427, i32 434, i32 441>
+  %strided.vec1 = shufflevector <448 x i64> %wide.vec, <448 x i64> poison, <64 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50, i32 57, i32 64, i32 71, i32 78, i32 85, i32 92, i32 99, i32 106, i32 113, i32 120, i32 127, i32 134, i32 141, i32 148, i32 155, i32 162, i32 169, i32 176, i32 183, i32 190, i32 197, i32 204, i32 211, i32 218, i32 225, i32 232, i32 239, i32 246, i32 253, i32 260, i32 267, i32 274, i32 281, i32 288, i32 295, i32 302, i32 309, i32 316, i32 323, i32 330, i32 337, i32 344, i32 351, i32 358, i32 365, i32 372, i32 379, i32 386, i32 393, i32 400, i32 407, i32 414, i32 421, i32 428, i32 435, i32 442>
+  %strided.vec2 = shufflevector <448 x i64> %wide.vec, <448 x i64> poison, <64 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51, i32 58, i32 65, i32 72, i32 79, i32 86, i32 93, i32 100, i32 107, i32 114, i32 121, i32 128, i32 135, i32 142, i32 149, i32 156, i32 163, i32 170, i32 177, i32 184, i32 191, i32 198, i32 205, i32 212, i32 219, i32 226, i32 233, i32 240, i32 247, i32 254, i32 261, i32 268, i32 275, i32 282, i32 289, i32 296, i32 303, i32 310, i32 317, i32 324, i32 331, i32 338, i32 345, i32 352, i32 359, i32 366, i32 373, i32 380, i32 387, i32 394, i32 401, i32 408, i32 415, i32 422, i32 429, i32 436, i32 443>
+  %strided.vec3 = shufflevector <448 x i64> %wide.vec, <448 x i64> poison, <64 x i32> <i32 3, i32 10, i32 17, i32 24, i32 31, i32 38, i32 45, i32 52, i32 59, i32 66, i32 73, i32 80, i32 87, i32 94, i32 101, i32 108, i32 115, i32 122, i32 129, i32 136, i32 143, i32 150, i32 157, i32 164, i32 171, i32 178, i32 185, i32 192, i32 199, i32 206, i32 213, i32 220, i32 227, i32 234, i32 241, i32 248, i32 255, i32 262, i32 269, i32 276, i32 283, i32 290, i32 297, i32 304, i32 311, i32 318, i32 325, i32 332, i32 339, i32 346, i32 353, i32 360, i32 367, i32 374, i32 381, i32 388, i32 395, i32 402, i32 409, i32 416, i32 423, i32 430, i32 437, i32 444>
+  %strided.vec4 = shufflevector <448 x i64> %wide.vec, <448 x i64> poison, <64 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53, i32 60, i32 67, i32 74, i32 81, i32 88, i32 95, i32 102, i32 109, i32 116, i32 123, i32 130, i32 137, i32 144, i32 151, i32 158, i32 165, i32 172, i32 179, i32 186, i32 193, i32 200, i32 207, i32 214, i32 221, i32 228, i32 235, i32 242, i32 249, i32 256, i32 263, i32 270, i32 277, i32 284, i32 291, i32 298, i32 305, i32 312, i32 319, i32 326, i32 333, i32 340, i32 347, i32 354, i32 361, i32 368, i32 375, i32 382, i32 389, i32 396, i32 403, i32 410, i32 417, i32 424, i32 431, i32 438, i32 445>
+  %strided.vec5 = shufflevector <448 x i64> %wide.vec, <448 x i64> poison, <64 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54, i32 61, i32 68, i32 75, i32 82, i32 89, i32 96, i32 103, i32 110, i32 117, i32 124, i32 131, i32 138, i32 145, i32 152, i32 159, i32 166, i32 173, i32 180, i32 187, i32 194, i32 201, i32 208, i32 215, i32 222, i32 229, i32 236, i32 243, i32 250, i32 257, i32 264, i32 271, i32 278, i32 285, i32 292, i32 299, i32 306, i32 313, i32 320, i32 327, i32 334, i32 341, i32 348, i32 355, i32 362, i32 369, i32 376, i32 383, i32 390, i32 397, i32 404, i32 411, i32 418, i32 425, i32 432, i32 439, i32 446>
+  %strided.vec6 = shufflevector <448 x i64> %wide.vec, <448 x i64> poison, <64 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55, i32 62, i32 69, i32 76, i32 83, i32 90, i32 97, i32 104, i32 111, i32 118, i32 125, i32 132, i32 139, i32 146, i32 153, i32 160, i32 167, i32 174, i32 181, i32 188, i32 195, i32 202, i32 209, i32 216, i32 223, i32 230, i32 237, i32 244, i32 251, i32 258, i32 265, i32 272, i32 279, i32 286, i32 293, i32 300, i32 307, i32 314, i32 321, i32 328, i32 335, i32 342, i32 349, i32 356, i32 363, i32 370, i32 377, i32 384, i32 391, i32 398, i32 405, i32 412, i32 419, i32 426, i32 433, i32 440, i32 447>
+  store <64 x i64> %strided.vec0, ptr %out.vec0, align 64
+  store <64 x i64> %strided.vec1, ptr %out.vec1, align 64
+  store <64 x i64> %strided.vec2, ptr %out.vec2, align 64
+  store <64 x i64> %strided.vec3, ptr %out.vec3, align 64
+  store <64 x i64> %strided.vec4, ptr %out.vec4, align 64
+  store <64 x i64> %strided.vec5, ptr %out.vec5, align 64
+  store <64 x i64> %strided.vec6, ptr %out.vec6, align 64
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-8.ll
index b9d7588..e7067d4 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-8.ll
@@ -145,7 +145,6 @@ define void @load_i64_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vmovaps %xmm3, (%rax)
 ; AVX512-NEXT:    retq
   %wide.vec = load <16 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <2 x i32> <i32 0, i32 8>
   %strided.vec1 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <2 x i32> <i32 1, i32 9>
   %strided.vec2 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <2 x i32> <i32 2, i32 10>
@@ -154,7 +153,6 @@ define void @load_i64_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <2 x i32> <i32 5, i32 13>
   %strided.vec6 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <2 x i32> <i32 6, i32 14>
   %strided.vec7 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <2 x i32> <i32 7, i32 15>
-
   store <2 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i64> %strided.vec2, ptr %out.vec2, align 64
@@ -163,7 +161,6 @@ define void @load_i64_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <2 x i64> %strided.vec5, ptr %out.vec5, align 64
   store <2 x i64> %strided.vec6, ptr %out.vec6, align 64
   store <2 x i64> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -407,7 +404,6 @@ define void @load_i64_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <32 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
   %strided.vec1 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <4 x i32> <i32 1, i32 9, i32 17, i32 25>
   %strided.vec2 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <4 x i32> <i32 2, i32 10, i32 18, i32 26>
@@ -416,7 +412,6 @@ define void @load_i64_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <4 x i32> <i32 5, i32 13, i32 21, i32 29>
   %strided.vec6 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <4 x i32> <i32 6, i32 14, i32 22, i32 30>
   %strided.vec7 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <4 x i32> <i32 7, i32 15, i32 23, i32 31>
-
   store <4 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i64> %strided.vec2, ptr %out.vec2, align 64
@@ -425,7 +420,6 @@ define void @load_i64_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <4 x i64> %strided.vec5, ptr %out.vec5, align 64
   store <4 x i64> %strided.vec6, ptr %out.vec6, align 64
   store <4 x i64> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -1085,7 +1079,6 @@ define void @load_i64_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <64 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <8 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56>
   %strided.vec1 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <8 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57>
   %strided.vec2 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <8 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58>
@@ -1094,7 +1087,6 @@ define void @load_i64_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <8 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61>
   %strided.vec6 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <8 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62>
   %strided.vec7 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <8 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63>
-
   store <8 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i64> %strided.vec2, ptr %out.vec2, align 64
@@ -1103,7 +1095,6 @@ define void @load_i64_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <8 x i64> %strided.vec5, ptr %out.vec5, align 64
   store <8 x i64> %strided.vec6, ptr %out.vec6, align 64
   store <8 x i64> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -2483,7 +2474,6 @@ define void @load_i64_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <128 x i64>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <16 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120>
   %strided.vec1 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <16 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121>
   %strided.vec2 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <16 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122>
@@ -2492,7 +2482,6 @@ define void @load_i64_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   %strided.vec5 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <16 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125>
   %strided.vec6 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <16 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126>
   %strided.vec7 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <16 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127>
-
   store <16 x i64> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i64> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i64> %strided.vec2, ptr %out.vec2, align 64
@@ -2501,7 +2490,9434 @@ define void @load_i64_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
   store <16 x i64> %strided.vec5, ptr %out.vec5, align 64
   store <16 x i64> %strided.vec6, ptr %out.vec6, align 64
   store <16 x i64> %strided.vec7, ptr %out.vec7, align 64
+  ret void
+}
 
+define void @load_i64_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
+; SSE-LABEL: load_i64_stride8_vf32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    subq $1688, %rsp # imm = 0x698
+; SSE-NEXT:    movaps 832(%rdi), %xmm0
+; SSE-NEXT:    movaps 320(%rdi), %xmm1
+; SSE-NEXT:    movaps 256(%rdi), %xmm8
+; SSE-NEXT:    movaps 960(%rdi), %xmm2
+; SSE-NEXT:    movaps 896(%rdi), %xmm10
+; SSE-NEXT:    movaps 448(%rdi), %xmm3
+; SSE-NEXT:    movaps 384(%rdi), %xmm9
+; SSE-NEXT:    movaps 576(%rdi), %xmm4
+; SSE-NEXT:    movaps 512(%rdi), %xmm12
+; SSE-NEXT:    movaps 64(%rdi), %xmm5
+; SSE-NEXT:    movaps (%rdi), %xmm11
+; SSE-NEXT:    movaps 704(%rdi), %xmm6
+; SSE-NEXT:    movaps 640(%rdi), %xmm14
+; SSE-NEXT:    movaps 192(%rdi), %xmm7
+; SSE-NEXT:    movaps 128(%rdi), %xmm13
+; SSE-NEXT:    movaps %xmm13, %xmm15
+; SSE-NEXT:    movlhps {{.*#+}} xmm15 = xmm15[0],xmm7[0]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm7[1]
+; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm11, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm5[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm5[1]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm9, %xmm5
+; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm3[0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm3[1]
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm8, %xmm3
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm14, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm6[1]
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm12, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm4[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm10, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 768(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1216(%rdi), %xmm0
+; SSE-NEXT:    movaps 1152(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1088(%rdi), %xmm0
+; SSE-NEXT:    movaps 1024(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1472(%rdi), %xmm0
+; SSE-NEXT:    movaps 1408(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1344(%rdi), %xmm0
+; SSE-NEXT:    movaps 1280(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1728(%rdi), %xmm0
+; SSE-NEXT:    movaps 1664(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1600(%rdi), %xmm0
+; SSE-NEXT:    movaps 1536(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1984(%rdi), %xmm0
+; SSE-NEXT:    movaps 1920(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1856(%rdi), %xmm0
+; SSE-NEXT:    movaps 1792(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%rdi), %xmm0
+; SSE-NEXT:    movaps 16(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 208(%rdi), %xmm0
+; SSE-NEXT:    movaps 144(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 336(%rdi), %xmm0
+; SSE-NEXT:    movaps 272(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 464(%rdi), %xmm0
+; SSE-NEXT:    movaps 400(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 592(%rdi), %xmm0
+; SSE-NEXT:    movaps 528(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 720(%rdi), %xmm0
+; SSE-NEXT:    movaps 656(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 848(%rdi), %xmm0
+; SSE-NEXT:    movaps 784(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 976(%rdi), %xmm0
+; SSE-NEXT:    movaps 912(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1104(%rdi), %xmm0
+; SSE-NEXT:    movaps 1040(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1232(%rdi), %xmm0
+; SSE-NEXT:    movaps 1168(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1360(%rdi), %xmm0
+; SSE-NEXT:    movaps 1296(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1488(%rdi), %xmm0
+; SSE-NEXT:    movaps 1424(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1616(%rdi), %xmm0
+; SSE-NEXT:    movaps 1552(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1744(%rdi), %xmm0
+; SSE-NEXT:    movaps 1680(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1872(%rdi), %xmm0
+; SSE-NEXT:    movaps 1808(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2000(%rdi), %xmm0
+; SSE-NEXT:    movaps 1936(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 96(%rdi), %xmm0
+; SSE-NEXT:    movaps 32(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 224(%rdi), %xmm0
+; SSE-NEXT:    movaps 160(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 352(%rdi), %xmm0
+; SSE-NEXT:    movaps 288(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 480(%rdi), %xmm0
+; SSE-NEXT:    movaps 416(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 608(%rdi), %xmm0
+; SSE-NEXT:    movaps 544(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 736(%rdi), %xmm0
+; SSE-NEXT:    movaps 672(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 864(%rdi), %xmm0
+; SSE-NEXT:    movaps 800(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 992(%rdi), %xmm0
+; SSE-NEXT:    movaps 928(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1120(%rdi), %xmm0
+; SSE-NEXT:    movaps 1056(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1248(%rdi), %xmm0
+; SSE-NEXT:    movaps 1184(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1376(%rdi), %xmm0
+; SSE-NEXT:    movaps 1312(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1504(%rdi), %xmm0
+; SSE-NEXT:    movaps 1440(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1632(%rdi), %xmm0
+; SSE-NEXT:    movaps 1568(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1760(%rdi), %xmm0
+; SSE-NEXT:    movaps 1696(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1888(%rdi), %xmm0
+; SSE-NEXT:    movaps 1824(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2016(%rdi), %xmm0
+; SSE-NEXT:    movaps 1952(%rdi), %xmm14
+; SSE-NEXT:    movaps %xmm14, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm0[1]
+; SSE-NEXT:    movaps 112(%rdi), %xmm0
+; SSE-NEXT:    movaps 48(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 240(%rdi), %xmm0
+; SSE-NEXT:    movaps 176(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 368(%rdi), %xmm0
+; SSE-NEXT:    movaps 304(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 496(%rdi), %xmm0
+; SSE-NEXT:    movaps 432(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 624(%rdi), %xmm0
+; SSE-NEXT:    movaps 560(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 752(%rdi), %xmm0
+; SSE-NEXT:    movaps 688(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 880(%rdi), %xmm0
+; SSE-NEXT:    movaps 816(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1008(%rdi), %xmm0
+; SSE-NEXT:    movaps 944(%rdi), %xmm15
+; SSE-NEXT:    movaps %xmm15, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm0[1]
+; SSE-NEXT:    movaps 1136(%rdi), %xmm0
+; SSE-NEXT:    movaps 1072(%rdi), %xmm12
+; SSE-NEXT:    movaps %xmm12, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm0[1]
+; SSE-NEXT:    movaps 1264(%rdi), %xmm0
+; SSE-NEXT:    movaps 1200(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm0[1]
+; SSE-NEXT:    movaps 1392(%rdi), %xmm0
+; SSE-NEXT:    movaps 1328(%rdi), %xmm6
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm0[1]
+; SSE-NEXT:    movaps 1520(%rdi), %xmm0
+; SSE-NEXT:    movaps 1456(%rdi), %xmm8
+; SSE-NEXT:    movaps %xmm8, %xmm13
+; SSE-NEXT:    movlhps {{.*#+}} xmm13 = xmm13[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
+; SSE-NEXT:    movaps 1648(%rdi), %xmm0
+; SSE-NEXT:    movaps 1584(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm11
+; SSE-NEXT:    movlhps {{.*#+}} xmm11 = xmm11[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm0[1]
+; SSE-NEXT:    movaps 1776(%rdi), %xmm0
+; SSE-NEXT:    movaps 1712(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm9
+; SSE-NEXT:    movlhps {{.*#+}} xmm9 = xmm9[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT:    movaps 1904(%rdi), %xmm0
+; SSE-NEXT:    movaps 1840(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm5
+; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 2032(%rdi), %xmm0
+; SSE-NEXT:    movaps 1968(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm0[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r9)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r9)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm14, 240(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm4, 240(%rax)
+; SSE-NEXT:    movaps %xmm5, 224(%rax)
+; SSE-NEXT:    movaps %xmm9, 208(%rax)
+; SSE-NEXT:    movaps %xmm11, 192(%rax)
+; SSE-NEXT:    movaps %xmm13, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm3, 240(%rax)
+; SSE-NEXT:    movaps %xmm1, 224(%rax)
+; SSE-NEXT:    movaps %xmm2, 208(%rax)
+; SSE-NEXT:    movaps %xmm7, 192(%rax)
+; SSE-NEXT:    movaps %xmm8, 176(%rax)
+; SSE-NEXT:    movaps %xmm6, 160(%rax)
+; SSE-NEXT:    movaps %xmm10, 144(%rax)
+; SSE-NEXT:    movaps %xmm12, 128(%rax)
+; SSE-NEXT:    movaps %xmm15, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    addq $1688, %rsp # imm = 0x698
+; SSE-NEXT:    retq
+;
+; AVX1-ONLY-LABEL: load_i64_stride8_vf32:
+; AVX1-ONLY:       # %bb.0:
+; AVX1-ONLY-NEXT:    subq $2296, %rsp # imm = 0x8F8
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm7[0],xmm6[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm5[0],xmm4[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps 1024(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm11[0],xmm10[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm11[1],xmm10[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm9[0],xmm8[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1600(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm9[1],xmm8[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1728(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm3[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1472(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1280(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1984(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1920(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1856(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1792(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1728(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1616(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1552(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm13
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm13[0],ymm0[0],ymm13[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm11[0],xmm14[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 592(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 528(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm10[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %ymm15
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm15[0],ymm0[0],ymm15[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1104(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1040(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm7[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm8[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1984(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1920(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1872(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1808(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm8[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1472(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1360(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1296(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm9[0],ymm6[0],ymm9[2],ymm6[2]
+; AVX1-ONLY-NEXT:    vmovaps 848(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps 784(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm4[0],xmm5[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm8[0],ymm3[0],ymm8[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovaps 336(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 272(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm12 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm11 = xmm11[1],xmm14[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm10[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = xmm7[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = xmm7[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm8[1],ymm3[1],ymm8[3],ymm3[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm9[1],ymm6[1],ymm9[3],ymm6[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1504(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1312(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1632(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1760(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1888(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1824(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2016(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1952(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovaps 368(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 304(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm3[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovaps 624(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps 560(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm10[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vmovaps 880(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vmovaps 816(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm12 = xmm4[0],xmm11[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm0[0],ymm5[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1136(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vmovaps 1072(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm5[0],xmm12[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1504(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1392(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vmovaps 1328(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm6[0],xmm13[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1760(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm0[0],ymm7[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1648(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vmovaps 1584(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm7[0],xmm14[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2016(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1952(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm8[0],ymm0[0],ymm8[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1904(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vmovaps 1840(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm8[0],xmm15[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm2[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm10[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm11[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm12[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm6[1],xmm13[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm14[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm8[1],xmm15[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 240(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 224(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 160(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 48(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 176(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 112(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 208(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 192(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 128(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, (%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 144(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 80(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 16(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 224(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 240(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 160(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 176(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 112(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 48(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 192(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 208(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, (%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 16(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 80(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 128(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 144(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, (%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 224(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 160(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 96(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 32(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 192(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 128(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 64(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, (%r8)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 240(%r9)
+; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 224(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 48(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 160(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 176(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 96(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 112(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 32(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 16(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, (%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 192(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 208(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 144(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 64(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 80(%r9)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 240(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 208(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, (%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 16(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 80(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 176(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 144(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 112(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 48(%rax)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, (%rax)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm10, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm9, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%rax)
+; AVX1-ONLY-NEXT:    addq $2296, %rsp # imm = 0x8F8
+; AVX1-ONLY-NEXT:    vzeroupper
+; AVX1-ONLY-NEXT:    retq
+;
+; AVX2-ONLY-LABEL: load_i64_stride8_vf32:
+; AVX2-ONLY:       # %bb.0:
+; AVX2-ONLY-NEXT:    subq $2408, %rsp # imm = 0x968
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 704(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps 640(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vmovaps 1152(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm6[0],xmm4[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 576(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vmovaps 512(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm5[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1088(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovaps 1024(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm10[0],xmm9[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1600(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm6[1],xmm4[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm10[1],xmm9[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm8[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1728(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm7[1],xmm5[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1664(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm3[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm8[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 960(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 896(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 832(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 768(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1472(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1280(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1984(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1920(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1856(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1792(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1600(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1536(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1728(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1664(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm5[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm9 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm9[2,3],ymm8[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 576(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 512(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 704(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 640(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm12[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1088(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1024(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1216(%rdi), %ymm14
+; AVX2-ONLY-NEXT:    vmovaps 1152(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm15[0],ymm14[0],ymm15[2],ymm14[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1856(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1792(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1984(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1920(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm4[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1344(%rdi), %ymm12
+; AVX2-ONLY-NEXT:    vmovaps 1280(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vmovaps 1472(%rdi), %ymm10
+; AVX2-ONLY-NEXT:    vmovaps 1408(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm8[0],ymm10[0],ymm8[2],ymm10[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm6[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 832(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vmovaps 768(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovaps 960(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovaps 896(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm9 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm9[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm9[0],ymm2[0],ymm9[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm13[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm13[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm13[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm15[1],ymm14[1],ymm15[3],ymm14[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm13[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm13[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm9[1],ymm2[1],ymm9[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm6[1],ymm7[1],ymm6[3],ymm7[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm8[1],ymm10[1],ymm8[3],ymm10[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 992(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 928(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 864(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 800(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1120(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1248(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1504(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1440(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1312(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 736(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 672(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 608(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 544(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1632(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1760(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1696(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1888(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1824(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2016(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1952(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm2[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 608(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 544(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 736(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 672(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 864(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 800(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 992(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 928(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1120(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1056(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm15
+; AVX2-ONLY-NEXT:    vmovaps 1184(%rdi), %ymm14
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm14[0],ymm15[0],ymm14[2],ymm15[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1376(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vmovaps 1312(%rdi), %ymm12
+; AVX2-ONLY-NEXT:    vmovaps 1504(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm10
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm10[0],ymm11[0],ymm10[2],ymm11[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm12[0],ymm13[0],ymm12[2],ymm13[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1632(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovaps 1568(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovaps 1760(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovaps 1696(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm5[0],ymm6[0],ymm5[2],ymm6[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm8[0],ymm9[0],ymm8[2],ymm9[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1888(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovaps 1824(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovaps 2016(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 1952(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = ymm4[1],mem[1],ymm4[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = ymm4[1],mem[1],ymm4[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm7[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = ymm4[1],mem[1],ymm4[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = ymm4[1],mem[1],ymm4[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = ymm4[1],mem[1],ymm4[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm14[1],ymm15[1],ymm14[3],ymm15[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm10[1],ymm11[1],ymm10[3],ymm11[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm7 = ymm12[1],ymm13[1],ymm12[3],ymm13[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm7[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm6[1],ymm5[3],ymm6[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm8[1],ymm9[1],ymm8[3],ymm9[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 240(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 224(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 32(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 160(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 96(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 48(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 176(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 112(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 208(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 192(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 128(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 64(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, (%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 144(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 80(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 16(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 224(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 240(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 160(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 176(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 96(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 112(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 32(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 48(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 192(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 208(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, (%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 16(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 64(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 80(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 128(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 144(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%r8)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 240(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 224(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 48(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 160(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 176(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 96(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 112(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 32(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 16(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, (%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 192(%r9)
+; AVX2-ONLY-NEXT:    vmovaps (%rsp), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 208(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 128(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 144(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 64(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 80(%r9)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 240(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 208(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, (%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 16(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 80(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 160(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 176(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 144(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 112(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 48(%rax)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rax)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm10, 160(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm14, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, (%rax)
+; AVX2-ONLY-NEXT:    addq $2408, %rsp # imm = 0x968
+; AVX2-ONLY-NEXT:    vzeroupper
+; AVX2-ONLY-NEXT:    retq
+;
+; AVX512F-LABEL: load_i64_stride8_vf32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    subq $2568, %rsp # imm = 0xA08
+; AVX512F-NEXT:    vmovdqa64 1856(%rdi), %zmm21
+; AVX512F-NEXT:    vmovdqa64 1984(%rdi), %zmm24
+; AVX512F-NEXT:    vmovdqa64 1344(%rdi), %zmm19
+; AVX512F-NEXT:    vmovdqa64 1280(%rdi), %zmm31
+; AVX512F-NEXT:    vmovdqa64 1472(%rdi), %zmm26
+; AVX512F-NEXT:    vmovdqa64 1408(%rdi), %zmm27
+; AVX512F-NEXT:    vmovdqa64 832(%rdi), %zmm22
+; AVX512F-NEXT:    vmovdqa64 768(%rdi), %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 960(%rdi), %zmm30
+; AVX512F-NEXT:    vmovdqa64 896(%rdi), %zmm10
+; AVX512F-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 320(%rdi), %zmm12
+; AVX512F-NEXT:    vmovdqa64 256(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 448(%rdi), %zmm20
+; AVX512F-NEXT:    vmovdqa64 384(%rdi), %zmm29
+; AVX512F-NEXT:    movb $-64, %al
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [2,10,2,10,2,10,2,10]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqa64 1216(%rdi), %ymm28
+; AVX512F-NEXT:    vmovdqa 1152(%rdi), %ymm1
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm4 = ymm1[0],ymm28[0],ymm1[2],ymm28[2]
+; AVX512F-NEXT:    vmovdqa64 1088(%rdi), %ymm25
+; AVX512F-NEXT:    vmovdqa 1024(%rdi), %ymm5
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm7 = ymm5[0],ymm25[0],ymm5[2],ymm25[2]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm4 = ymm7[2,3],ymm4[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm4, %zmm2, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm22, %zmm0, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm4 {%k1}
+; AVX512F-NEXT:    vmovdqa64 704(%rdi), %ymm23
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %ymm17
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm8 = ymm17[0],ymm23[0],ymm17[2],ymm23[2]
+; AVX512F-NEXT:    vmovdqa 576(%rdi), %ymm2
+; AVX512F-NEXT:    vmovdqa 512(%rdi), %ymm10
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm11 = ymm10[0],ymm2[0],ymm10[2],ymm2[2]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm8 = ymm11[2,3],ymm8[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm8, %zmm4, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm29, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm20, %zmm0, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm7
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm8 {%k1}
+; AVX512F-NEXT:    vmovdqa 192(%rdi), %ymm9
+; AVX512F-NEXT:    vmovdqa 128(%rdi), %ymm11
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm12 = ymm11[0],ymm9[0],ymm11[2],ymm9[2]
+; AVX512F-NEXT:    vmovdqa 64(%rdi), %ymm4
+; AVX512F-NEXT:    vmovdqa64 (%rdi), %ymm16
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm15 = ymm16[0],ymm4[0],ymm16[2],ymm4[2]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm12 = ymm15[2,3],ymm12[2,3]
+; AVX512F-NEXT:    vmovdqa64 1920(%rdi), %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm12, %zmm8, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm8
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqa64 1792(%rdi), %zmm12
+; AVX512F-NEXT:    vmovdqu64 %zmm12, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm21, %zmm12, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm0 {%k1}
+; AVX512F-NEXT:    vmovdqa 1728(%rdi), %ymm14
+; AVX512F-NEXT:    vmovdqa 1664(%rdi), %ymm12
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm15 = ymm12[0],ymm14[0],ymm12[2],ymm14[2]
+; AVX512F-NEXT:    vmovdqa64 1600(%rdi), %ymm18
+; AVX512F-NEXT:    vmovdqa 1536(%rdi), %ymm8
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm13 = ymm8[0],ymm18[0],ymm8[2],ymm18[2]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm13[2,3],ymm15[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm13, %zmm0, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [3,11,3,11,3,11,3,11]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm13
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm0, %zmm13
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm15
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm0, %zmm15
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm15 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm1 = ymm1[1],ymm28[1],ymm1[3],ymm28[3]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm3 = ymm5[1],ymm25[1],ymm5[3],ymm25[3]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm3[2,3],ymm1[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm15, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm13
+; AVX512F-NEXT:    vpermt2q %zmm22, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm3 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm1 = ymm17[1],ymm23[1],ymm17[3],ymm23[3]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm2 = ymm10[1],ymm2[1],ymm10[3],ymm2[3]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 %zmm29, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm29, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm20, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm20, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm1 = ymm11[1],ymm9[1],ymm11[3],ymm9[3]
+; AVX512F-NEXT:    vmovdqa64 1216(%rdi), %zmm9
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm3 = ymm16[1],ymm4[1],ymm16[3],ymm4[3]
+; AVX512F-NEXT:    vmovdqa64 1152(%rdi), %zmm10
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm3[2,3],ymm1[2,3]
+; AVX512F-NEXT:    vmovdqa64 1088(%rdi), %zmm11
+; AVX512F-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm23 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2q %zmm21, %zmm23, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm0 {%k1}
+; AVX512F-NEXT:    vmovdqa64 1024(%rdi), %zmm4
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm1 = ymm12[1],ymm14[1],ymm12[3],ymm14[3]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm2 = ymm8[1],ymm18[1],ymm8[3],ymm18[3]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [4,12,4,12,4,12,4,12]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm0, %zmm1
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm2 = zmm31[0],zmm19[0],zmm31[2],zmm19[2],zmm31[4],zmm19[4],zmm31[6],zmm19[6]
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm12
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm11, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,4,12>
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm4
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm1, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm3, %zmm2, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm0, %zmm2
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm15[0],zmm13[0],zmm15[2],zmm13[2],zmm15[4],zmm13[4],zmm15[6],zmm13[6]
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm11
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm16
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 512(%rdi), %zmm13
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 704(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm1, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm20, %zmm0, %zmm29
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm6[0],zmm7[0],zmm6[2],zmm7[2],zmm6[4],zmm7[4],zmm6[6],zmm7[6]
+; AVX512F-NEXT:    vmovdqa64 %zmm29, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqa64 64(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 (%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 192(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 128(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm1, %zmm4
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm4
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm23[0],zmm21[0],zmm23[2],zmm21[2],zmm23[4],zmm21[4],zmm23[6],zmm21[6]
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm6
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqa64 1600(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1536(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm7, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqa64 1728(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1664(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm7, %zmm2, %zmm1
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm18 = [0,8,0,8,0,8,0,8]
+; AVX512F-NEXT:    # zmm18 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm18, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm23 = [1,9,1,9,1,9,1,9]
+; AVX512F-NEXT:    # zmm23 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm23, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm3 = [6,14,6,14,6,14,6,14]
+; AVX512F-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm8 = zmm31[1],zmm19[1],zmm31[3],zmm19[3],zmm31[5],zmm19[5],zmm31[7],zmm19[7]
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm10 = [7,15,7,15,7,15,7,15]
+; AVX512F-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm10, %zmm31
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm1 {%k1} = zmm27[0],zmm26[0],zmm27[2],zmm26[2],zmm27[4],zmm26[4],zmm27[6],zmm26[6]
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm31 {%k1} = zmm27[1],zmm26[1],zmm27[3],zmm26[3],zmm27[5],zmm26[5],zmm27[7],zmm26[7]
+; AVX512F-NEXT:    vmovdqu64 %zmm31, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm7
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm18, %zmm27
+; AVX512F-NEXT:    vmovdqu64 %zmm27, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm23, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [5,13,5,13,5,13,5,13]
+; AVX512F-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm8 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm2, %zmm12
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm7 = <u,u,5,13>
+; AVX512F-NEXT:    vmovdqa64 %zmm14, %zmm15
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm7, %zmm15
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm12[0,1,2,3],ymm15[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm8, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm27
+; AVX512F-NEXT:    vpermt2q %zmm16, %zmm18, %zmm27
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm22
+; AVX512F-NEXT:    vpermt2q %zmm16, %zmm23, %zmm22
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm21
+; AVX512F-NEXT:    vpermt2q %zmm16, %zmm3, %zmm21
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm14 = zmm11[1],zmm16[1],zmm11[3],zmm16[3],zmm11[5],zmm16[5],zmm11[7],zmm16[7]
+; AVX512F-NEXT:    vpermt2q %zmm16, %zmm10, %zmm11
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm21 {%k1} = zmm16[0],zmm30[0],zmm16[2],zmm30[2],zmm16[4],zmm30[4],zmm16[6],zmm30[6]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm11 {%k1} = zmm16[1],zmm30[1],zmm16[3],zmm30[3],zmm16[5],zmm30[5],zmm16[7],zmm30[7]
+; AVX512F-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm29
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm11
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm18, %zmm16
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm23, %zmm29
+; AVX512F-NEXT:    vpermt2q %zmm30, %zmm2, %zmm11
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm8
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm2, %zmm8
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm7, %zmm15
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm15[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm31
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm18, %zmm31
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm20
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm23, %zmm20
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm26
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm3, %zmm26
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm9 = zmm5[1],zmm1[1],zmm5[3],zmm1[3],zmm5[5],zmm1[5],zmm5[7],zmm1[7]
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm10, %zmm5
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm26 {%k1} = zmm12[0],zmm28[0],zmm12[2],zmm28[2],zmm12[4],zmm28[4],zmm12[6],zmm28[6]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm5 {%k1} = zmm12[1],zmm28[1],zmm12[3],zmm28[3],zmm12[5],zmm28[5],zmm12[7],zmm28[7]
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm24
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm5
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm18, %zmm12
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm23, %zmm24
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm2, %zmm5
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm25
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm18, %zmm25
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm1
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2q %zmm1, %zmm6, %zmm18
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm28
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm23, %zmm28
+; AVX512F-NEXT:    vpermi2q %zmm1, %zmm6, %zmm23
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm17
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm3, %zmm17
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm8 = zmm6[1],zmm1[1],zmm6[3],zmm1[3],zmm6[5],zmm1[5],zmm6[7],zmm1[7]
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm10, %zmm6
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm17 {%k1} = zmm30[0],zmm4[0],zmm30[2],zmm4[2],zmm30[4],zmm4[4],zmm30[6],zmm4[6]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm6 {%k1} = zmm30[1],zmm4[1],zmm30[3],zmm4[3],zmm30[5],zmm4[5],zmm30[7],zmm4[7]
+; AVX512F-NEXT:    vmovdqu64 %zmm6, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm2, %zmm30
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm6
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm15
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm3, %zmm6
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm10, %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm2, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm0, %zmm10, %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm4
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm15, %zmm4
+; AVX512F-NEXT:    vpermi2q %zmm1, %zmm0, %zmm15
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm10, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm10, %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm14 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm14, %zmm0 # 32-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm9 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm5
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm13, %zmm7, %zmm5
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm5[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm9, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm8 {%k1}
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2q %zmm9, %zmm5, %zmm7
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm7[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm8, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm0 = <u,u,6,14>
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm7 = <u,u,7,15>
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm7, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm11
+; AVX512F-NEXT:    vpermt2q %zmm13, %zmm0, %zmm11
+; AVX512F-NEXT:    vpermt2q %zmm13, %zmm7, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm10
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm10
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm2
+; AVX512F-NEXT:    vpermi2q %zmm9, %zmm5, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm7, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm7, %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm10[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm4, %zmm1, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm3, %zmm21, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm31 {%k1}
+; AVX512F-NEXT:    vmovdqa 192(%rdi), %xmm14
+; AVX512F-NEXT:    vmovdqa 128(%rdi), %xmm13
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm8 = xmm13[0],xmm14[0]
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm10
+; AVX512F-NEXT:    vmovdqa 64(%rdi), %xmm7
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm21 = xmm10[0],xmm7[0]
+; AVX512F-NEXT:    vinserti32x4 $1, %xmm8, %ymm21, %ymm8
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm8, %zmm31, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm27 {%k1}
+; AVX512F-NEXT:    vmovdqa 704(%rdi), %xmm12
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %xmm21
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm30 = xmm21[0],xmm12[0]
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %xmm31
+; AVX512F-NEXT:    vmovdqa 512(%rdi), %xmm5
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm9 = xmm5[0],xmm31[0]
+; AVX512F-NEXT:    vinserti32x4 $1, %xmm30, %ymm9, %ymm9
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm9, %zmm27, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqa64 1216(%rdi), %xmm19
+; AVX512F-NEXT:    vmovdqa64 1152(%rdi), %xmm27
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm30 = xmm27[0],xmm19[0]
+; AVX512F-NEXT:    vmovdqa 1088(%rdi), %xmm2
+; AVX512F-NEXT:    vmovdqa 1024(%rdi), %xmm1
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm4 = xmm1[0],xmm2[0]
+; AVX512F-NEXT:    vinserti32x4 $1, %xmm30, %ymm4, %ymm4
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm4, %zmm3, %zmm8
+; AVX512F-NEXT:    vmovdqa64 %zmm25, %zmm18 {%k1}
+; AVX512F-NEXT:    vmovdqa64 1728(%rdi), %xmm25
+; AVX512F-NEXT:    vmovdqa64 1664(%rdi), %xmm30
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm6 = xmm30[0],xmm25[0]
+; AVX512F-NEXT:    vmovdqa 1600(%rdi), %xmm4
+; AVX512F-NEXT:    vmovdqa 1536(%rdi), %xmm3
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm9 = xmm3[0],xmm4[0]
+; AVX512F-NEXT:    vinserti128 $1, %xmm6, %ymm9, %ymm6
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm6, %zmm18, %zmm6
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm16 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm9 = xmm27[1],xmm19[1]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX512F-NEXT:    vinserti128 $1, %xmm9, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm1, %zmm16, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm29, %zmm22 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm21[1],xmm12[1]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm5 = xmm5[1],xmm31[1]
+; AVX512F-NEXT:    vinserti128 $1, %xmm2, %ymm5, %ymm2
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm22, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm20 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm5 = xmm13[1],xmm14[1]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm7 = xmm10[1],xmm7[1]
+; AVX512F-NEXT:    vinserti128 $1, %xmm5, %ymm7, %ymm5
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm5, %zmm20, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm28, %zmm23 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm7 = xmm30[1],xmm25[1]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
+; AVX512F-NEXT:    vinserti128 $1, %xmm7, %ymm3, %ymm3
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm3, %zmm23, %zmm3
+; AVX512F-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm4 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm4 = mem[0,1,2,3],ymm11[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm4, %zmm26, %zmm4
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm17, %zmm0
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm7 = mem[0,1,2,3],ymm7[4,5,6,7]
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vinsertf64x4 $0, %ymm7, %zmm9, %zmm7
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm9 = mem[0,1,2,3],ymm9[4,5,6,7]
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vinsertf64x4 $0, %ymm9, %zmm10, %zmm9
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm10 = mem[0,1,2,3],ymm10[4,5,6,7]
+; AVX512F-NEXT:    vmovups (%rsp), %zmm11 # 64-byte Reload
+; AVX512F-NEXT:    vinsertf64x4 $0, %ymm10, %zmm11, %zmm10
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512F-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm11 = mem[0,1,2,3],ymm11[4,5,6,7]
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vinsertf64x4 $0, %ymm11, %zmm12, %zmm11
+; AVX512F-NEXT:    vmovdqa64 %zmm6, 192(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm8, 128(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm6, 64(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm6, (%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm3, 192(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm5, (%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm2, 64(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm1, 128(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, 192(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, (%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, 64(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, 128(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, 192(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, (%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, 64(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, 128(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, 192(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, (%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, 64(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, 128(%r9)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, 192(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, (%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, 64(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm1, 128(%rax)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovdqa64 %zmm0, 192(%rax)
+; AVX512F-NEXT:    vmovdqa64 %zmm4, (%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%rax)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovaps %zmm11, 128(%rax)
+; AVX512F-NEXT:    vmovaps %zmm10, 192(%rax)
+; AVX512F-NEXT:    vmovaps %zmm9, (%rax)
+; AVX512F-NEXT:    vmovaps %zmm7, 64(%rax)
+; AVX512F-NEXT:    addq $2568, %rsp # imm = 0xA08
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: load_i64_stride8_vf32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    subq $2568, %rsp # imm = 0xA08
+; AVX512BW-NEXT:    vmovdqa64 1856(%rdi), %zmm21
+; AVX512BW-NEXT:    vmovdqa64 1984(%rdi), %zmm24
+; AVX512BW-NEXT:    vmovdqa64 1344(%rdi), %zmm19
+; AVX512BW-NEXT:    vmovdqa64 1280(%rdi), %zmm31
+; AVX512BW-NEXT:    vmovdqa64 1472(%rdi), %zmm26
+; AVX512BW-NEXT:    vmovdqa64 1408(%rdi), %zmm27
+; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm22
+; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm30
+; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm10
+; AVX512BW-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm12
+; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm20
+; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm29
+; AVX512BW-NEXT:    movb $-64, %al
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [2,10,2,10,2,10,2,10]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %ymm28
+; AVX512BW-NEXT:    vmovdqa 1152(%rdi), %ymm1
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm4 = ymm1[0],ymm28[0],ymm1[2],ymm28[2]
+; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %ymm25
+; AVX512BW-NEXT:    vmovdqa 1024(%rdi), %ymm5
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm7 = ymm5[0],ymm25[0],ymm5[2],ymm25[2]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm4 = ymm7[2,3],ymm4[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm4, %zmm2, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm4 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %ymm23
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %ymm17
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm8 = ymm17[0],ymm23[0],ymm17[2],ymm23[2]
+; AVX512BW-NEXT:    vmovdqa 576(%rdi), %ymm2
+; AVX512BW-NEXT:    vmovdqa 512(%rdi), %ymm10
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm11 = ymm10[0],ymm2[0],ymm10[2],ymm2[2]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm8 = ymm11[2,3],ymm8[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm8, %zmm4, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm20, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm7
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm8 {%k1}
+; AVX512BW-NEXT:    vmovdqa 192(%rdi), %ymm9
+; AVX512BW-NEXT:    vmovdqa 128(%rdi), %ymm11
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm12 = ymm11[0],ymm9[0],ymm11[2],ymm9[2]
+; AVX512BW-NEXT:    vmovdqa 64(%rdi), %ymm4
+; AVX512BW-NEXT:    vmovdqa64 (%rdi), %ymm16
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm15 = ymm16[0],ymm4[0],ymm16[2],ymm4[2]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm12 = ymm15[2,3],ymm12[2,3]
+; AVX512BW-NEXT:    vmovdqa64 1920(%rdi), %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm12, %zmm8, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm8
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 1792(%rdi), %zmm12
+; AVX512BW-NEXT:    vmovdqu64 %zmm12, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm21, %zmm12, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm0 {%k1}
+; AVX512BW-NEXT:    vmovdqa 1728(%rdi), %ymm14
+; AVX512BW-NEXT:    vmovdqa 1664(%rdi), %ymm12
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm15 = ymm12[0],ymm14[0],ymm12[2],ymm14[2]
+; AVX512BW-NEXT:    vmovdqa64 1600(%rdi), %ymm18
+; AVX512BW-NEXT:    vmovdqa 1536(%rdi), %ymm8
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm13 = ymm8[0],ymm18[0],ymm8[2],ymm18[2]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm13[2,3],ymm15[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm13, %zmm0, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [3,11,3,11,3,11,3,11]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm13
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm0, %zmm13
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm15
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm0, %zmm15
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm15 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm1 = ymm1[1],ymm28[1],ymm1[3],ymm28[3]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm3 = ymm5[1],ymm25[1],ymm5[3],ymm25[3]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm3[2,3],ymm1[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm15, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm13
+; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm3 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm1 = ymm17[1],ymm23[1],ymm17[3],ymm23[3]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm2 = ymm10[1],ymm2[1],ymm10[3],ymm2[3]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm3, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 %zmm29, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm20, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm20, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm1 = ymm11[1],ymm9[1],ymm11[3],ymm9[3]
+; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %zmm9
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm3 = ymm16[1],ymm4[1],ymm16[3],ymm4[3]
+; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %zmm10
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm3[2,3],ymm1[2,3]
+; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %zmm11
+; AVX512BW-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm23 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm21, %zmm23, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %zmm4
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm1 = ymm12[1],ymm14[1],ymm12[3],ymm14[3]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm2 = ymm8[1],ymm18[1],ymm8[3],ymm18[3]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [4,12,4,12,4,12,4,12]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm2 = zmm31[0],zmm19[0],zmm31[2],zmm19[2],zmm31[4],zmm19[4],zmm31[6],zmm19[6]
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm12
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,4,12>
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm4
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm1, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm3, %zmm2, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm0, %zmm2
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm15[0],zmm13[0],zmm15[2],zmm13[2],zmm15[4],zmm13[4],zmm15[6],zmm13[6]
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm11
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm16
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm13
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm5
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm1, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm20, %zmm0, %zmm29
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm6[0],zmm7[0],zmm6[2],zmm7[2],zmm6[4],zmm7[4],zmm6[6],zmm7[6]
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm1, %zmm4
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm4
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm23[0],zmm21[0],zmm23[2],zmm21[2],zmm23[4],zmm21[4],zmm23[6],zmm21[6]
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm6
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 1600(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1536(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm7, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 1728(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1664(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm7, %zmm2, %zmm1
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm18 = [0,8,0,8,0,8,0,8]
+; AVX512BW-NEXT:    # zmm18 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm18, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm23 = [1,9,1,9,1,9,1,9]
+; AVX512BW-NEXT:    # zmm23 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm23, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm3 = [6,14,6,14,6,14,6,14]
+; AVX512BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm8 = zmm31[1],zmm19[1],zmm31[3],zmm19[3],zmm31[5],zmm19[5],zmm31[7],zmm19[7]
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm10 = [7,15,7,15,7,15,7,15]
+; AVX512BW-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm10, %zmm31
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm1 {%k1} = zmm27[0],zmm26[0],zmm27[2],zmm26[2],zmm27[4],zmm26[4],zmm27[6],zmm26[6]
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm31 {%k1} = zmm27[1],zmm26[1],zmm27[3],zmm26[3],zmm27[5],zmm26[5],zmm27[7],zmm26[7]
+; AVX512BW-NEXT:    vmovdqu64 %zmm31, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm18, %zmm27
+; AVX512BW-NEXT:    vmovdqu64 %zmm27, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm23, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [5,13,5,13,5,13,5,13]
+; AVX512BW-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm8 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm2, %zmm12
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm7 = <u,u,5,13>
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, %zmm15
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm7, %zmm15
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm12[0,1,2,3],ymm15[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm8, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm27
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm18, %zmm27
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm22
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm23, %zmm22
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm21
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm3, %zmm21
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm14 = zmm11[1],zmm16[1],zmm11[3],zmm16[3],zmm11[5],zmm16[5],zmm11[7],zmm16[7]
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm10, %zmm11
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm21 {%k1} = zmm16[0],zmm30[0],zmm16[2],zmm30[2],zmm16[4],zmm30[4],zmm16[6],zmm30[6]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm11 {%k1} = zmm16[1],zmm30[1],zmm16[3],zmm30[3],zmm16[5],zmm30[5],zmm16[7],zmm30[7]
+; AVX512BW-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm29
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm11
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm18, %zmm16
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm23, %zmm29
+; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm2, %zmm11
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm2, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm7, %zmm15
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm15[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm31
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm18, %zmm31
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm20
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm23, %zmm20
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm26
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm3, %zmm26
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm9 = zmm5[1],zmm1[1],zmm5[3],zmm1[3],zmm5[5],zmm1[5],zmm5[7],zmm1[7]
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm10, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm26 {%k1} = zmm12[0],zmm28[0],zmm12[2],zmm28[2],zmm12[4],zmm28[4],zmm12[6],zmm28[6]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm5 {%k1} = zmm12[1],zmm28[1],zmm12[3],zmm28[3],zmm12[5],zmm28[5],zmm12[7],zmm28[7]
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm24
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm5
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm18, %zmm12
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm23, %zmm24
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm2, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm25
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm18, %zmm25
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm6, %zmm18
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm28
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm23, %zmm28
+; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm6, %zmm23
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm17
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm3, %zmm17
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm8 = zmm6[1],zmm1[1],zmm6[3],zmm1[3],zmm6[5],zmm1[5],zmm6[7],zmm1[7]
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm10, %zmm6
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm17 {%k1} = zmm30[0],zmm4[0],zmm30[2],zmm4[2],zmm30[4],zmm4[4],zmm30[6],zmm4[6]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm6 {%k1} = zmm30[1],zmm4[1],zmm30[3],zmm4[3],zmm30[5],zmm4[5],zmm30[7],zmm4[7]
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm2, %zmm30
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm6
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm15
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm3, %zmm6
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm10, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm2, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm10, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm4
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm15, %zmm4
+; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm15
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm10, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm10, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm14 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm14, %zmm0 # 32-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm9 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm7, %zmm5
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm5[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm9, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm8 {%k1}
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm9, %zmm5, %zmm7
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm7[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm8, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm0 = <u,u,6,14>
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm7 = <u,u,7,15>
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm7, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm11
+; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm0, %zmm11
+; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm7, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm10
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm0, %zmm10
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm2
+; AVX512BW-NEXT:    vpermi2q %zmm9, %zmm5, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm7, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm7, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm10[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm4, %zmm1, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm3, %zmm21, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm31 {%k1}
+; AVX512BW-NEXT:    vmovdqa 192(%rdi), %xmm14
+; AVX512BW-NEXT:    vmovdqa 128(%rdi), %xmm13
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm8 = xmm13[0],xmm14[0]
+; AVX512BW-NEXT:    vmovdqa (%rdi), %xmm10
+; AVX512BW-NEXT:    vmovdqa 64(%rdi), %xmm7
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm21 = xmm10[0],xmm7[0]
+; AVX512BW-NEXT:    vinserti32x4 $1, %xmm8, %ymm21, %ymm8
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm8, %zmm31, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm27 {%k1}
+; AVX512BW-NEXT:    vmovdqa 704(%rdi), %xmm12
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %xmm21
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm30 = xmm21[0],xmm12[0]
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %xmm31
+; AVX512BW-NEXT:    vmovdqa 512(%rdi), %xmm5
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm9 = xmm5[0],xmm31[0]
+; AVX512BW-NEXT:    vinserti32x4 $1, %xmm30, %ymm9, %ymm9
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm9, %zmm27, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %xmm19
+; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %xmm27
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm30 = xmm27[0],xmm19[0]
+; AVX512BW-NEXT:    vmovdqa 1088(%rdi), %xmm2
+; AVX512BW-NEXT:    vmovdqa 1024(%rdi), %xmm1
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm4 = xmm1[0],xmm2[0]
+; AVX512BW-NEXT:    vinserti32x4 $1, %xmm30, %ymm4, %ymm4
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm4, %zmm3, %zmm8
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm18 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 1728(%rdi), %xmm25
+; AVX512BW-NEXT:    vmovdqa64 1664(%rdi), %xmm30
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm6 = xmm30[0],xmm25[0]
+; AVX512BW-NEXT:    vmovdqa 1600(%rdi), %xmm4
+; AVX512BW-NEXT:    vmovdqa 1536(%rdi), %xmm3
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm9 = xmm3[0],xmm4[0]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm6, %ymm9, %ymm6
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm6, %zmm18, %zmm6
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm16 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm9 = xmm27[1],xmm19[1]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm9, %ymm1, %ymm1
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm1, %zmm16, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm29, %zmm22 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm21[1],xmm12[1]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm5 = xmm5[1],xmm31[1]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm2, %ymm5, %ymm2
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm22, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm20 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm5 = xmm13[1],xmm14[1]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm7 = xmm10[1],xmm7[1]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm5, %ymm7, %ymm5
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm5, %zmm20, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm23 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm7 = xmm30[1],xmm25[1]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm7, %ymm3, %ymm3
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm3, %zmm23, %zmm3
+; AVX512BW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm4 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm4 = mem[0,1,2,3],ymm11[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm4, %zmm26, %zmm4
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm17, %zmm0
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm7 = mem[0,1,2,3],ymm7[4,5,6,7]
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vinsertf64x4 $0, %ymm7, %zmm9, %zmm7
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm9 = mem[0,1,2,3],ymm9[4,5,6,7]
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vinsertf64x4 $0, %ymm9, %zmm10, %zmm9
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm10 = mem[0,1,2,3],ymm10[4,5,6,7]
+; AVX512BW-NEXT:    vmovups (%rsp), %zmm11 # 64-byte Reload
+; AVX512BW-NEXT:    vinsertf64x4 $0, %ymm10, %zmm11, %zmm10
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512BW-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm11 = mem[0,1,2,3],ymm11[4,5,6,7]
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vinsertf64x4 $0, %ymm11, %zmm12, %zmm11
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, 192(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, 128(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm6, 64(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm6, (%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, 192(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, (%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, 64(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, 128(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, 192(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, (%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, 64(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, 128(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, 192(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, (%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, 64(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, 128(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, 192(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, (%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, 64(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, 128(%r9)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, 192(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, (%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, 64(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm1, 128(%rax)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, 192(%rax)
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, (%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%rax)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovaps %zmm11, 128(%rax)
+; AVX512BW-NEXT:    vmovaps %zmm10, 192(%rax)
+; AVX512BW-NEXT:    vmovaps %zmm9, (%rax)
+; AVX512BW-NEXT:    vmovaps %zmm7, 64(%rax)
+; AVX512BW-NEXT:    addq $2568, %rsp # imm = 0xA08
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %wide.vec = load <256 x i64>, ptr %in.vec, align 64
+  %strided.vec0 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <32 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120, i32 128, i32 136, i32 144, i32 152, i32 160, i32 168, i32 176, i32 184, i32 192, i32 200, i32 208, i32 216, i32 224, i32 232, i32 240, i32 248>
+  %strided.vec1 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <32 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121, i32 129, i32 137, i32 145, i32 153, i32 161, i32 169, i32 177, i32 185, i32 193, i32 201, i32 209, i32 217, i32 225, i32 233, i32 241, i32 249>
+  %strided.vec2 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <32 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122, i32 130, i32 138, i32 146, i32 154, i32 162, i32 170, i32 178, i32 186, i32 194, i32 202, i32 210, i32 218, i32 226, i32 234, i32 242, i32 250>
+  %strided.vec3 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <32 x i32> <i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 51, i32 59, i32 67, i32 75, i32 83, i32 91, i32 99, i32 107, i32 115, i32 123, i32 131, i32 139, i32 147, i32 155, i32 163, i32 171, i32 179, i32 187, i32 195, i32 203, i32 211, i32 219, i32 227, i32 235, i32 243, i32 251>
+  %strided.vec4 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <32 x i32> <i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 52, i32 60, i32 68, i32 76, i32 84, i32 92, i32 100, i32 108, i32 116, i32 124, i32 132, i32 140, i32 148, i32 156, i32 164, i32 172, i32 180, i32 188, i32 196, i32 204, i32 212, i32 220, i32 228, i32 236, i32 244, i32 252>
+  %strided.vec5 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <32 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125, i32 133, i32 141, i32 149, i32 157, i32 165, i32 173, i32 181, i32 189, i32 197, i32 205, i32 213, i32 221, i32 229, i32 237, i32 245, i32 253>
+  %strided.vec6 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <32 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126, i32 134, i32 142, i32 150, i32 158, i32 166, i32 174, i32 182, i32 190, i32 198, i32 206, i32 214, i32 222, i32 230, i32 238, i32 246, i32 254>
+  %strided.vec7 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <32 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127, i32 135, i32 143, i32 151, i32 159, i32 167, i32 175, i32 183, i32 191, i32 199, i32 207, i32 215, i32 223, i32 231, i32 239, i32 247, i32 255>
+  store <32 x i64> %strided.vec0, ptr %out.vec0, align 64
+  store <32 x i64> %strided.vec1, ptr %out.vec1, align 64
+  store <32 x i64> %strided.vec2, ptr %out.vec2, align 64
+  store <32 x i64> %strided.vec3, ptr %out.vec3, align 64
+  store <32 x i64> %strided.vec4, ptr %out.vec4, align 64
+  store <32 x i64> %strided.vec5, ptr %out.vec5, align 64
+  store <32 x i64> %strided.vec6, ptr %out.vec6, align 64
+  store <32 x i64> %strided.vec7, ptr %out.vec7, align 64
+  ret void
+}
+
+define void @load_i64_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
+; SSE-LABEL: load_i64_stride8_vf64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    subq $3720, %rsp # imm = 0xE88
+; SSE-NEXT:    movaps 960(%rdi), %xmm0
+; SSE-NEXT:    movaps 832(%rdi), %xmm1
+; SSE-NEXT:    movaps 768(%rdi), %xmm8
+; SSE-NEXT:    movaps 704(%rdi), %xmm2
+; SSE-NEXT:    movaps 640(%rdi), %xmm9
+; SSE-NEXT:    movaps 576(%rdi), %xmm3
+; SSE-NEXT:    movaps 512(%rdi), %xmm10
+; SSE-NEXT:    movaps 448(%rdi), %xmm4
+; SSE-NEXT:    movaps 384(%rdi), %xmm11
+; SSE-NEXT:    movaps 320(%rdi), %xmm5
+; SSE-NEXT:    movaps 256(%rdi), %xmm12
+; SSE-NEXT:    movaps 192(%rdi), %xmm6
+; SSE-NEXT:    movaps 128(%rdi), %xmm13
+; SSE-NEXT:    movaps 64(%rdi), %xmm7
+; SSE-NEXT:    movaps (%rdi), %xmm14
+; SSE-NEXT:    movaps %xmm14, %xmm15
+; SSE-NEXT:    movlhps {{.*#+}} xmm15 = xmm15[0],xmm7[0]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm7[1]
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm13, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm6[1]
+; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm12, %xmm6
+; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm5[0]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm5[1]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm11, %xmm5
+; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm10, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm3[1]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm9, %xmm3
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm8, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 896(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1088(%rdi), %xmm0
+; SSE-NEXT:    movaps 1024(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1216(%rdi), %xmm0
+; SSE-NEXT:    movaps 1152(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1344(%rdi), %xmm0
+; SSE-NEXT:    movaps 1280(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1472(%rdi), %xmm0
+; SSE-NEXT:    movaps 1408(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1600(%rdi), %xmm0
+; SSE-NEXT:    movaps 1536(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1728(%rdi), %xmm0
+; SSE-NEXT:    movaps 1664(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1856(%rdi), %xmm0
+; SSE-NEXT:    movaps 1792(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1984(%rdi), %xmm0
+; SSE-NEXT:    movaps 1920(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2112(%rdi), %xmm0
+; SSE-NEXT:    movaps 2048(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2240(%rdi), %xmm0
+; SSE-NEXT:    movaps 2176(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2368(%rdi), %xmm0
+; SSE-NEXT:    movaps 2304(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2496(%rdi), %xmm0
+; SSE-NEXT:    movaps 2432(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2624(%rdi), %xmm0
+; SSE-NEXT:    movaps 2560(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2752(%rdi), %xmm0
+; SSE-NEXT:    movaps 2688(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2880(%rdi), %xmm0
+; SSE-NEXT:    movaps 2816(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3008(%rdi), %xmm0
+; SSE-NEXT:    movaps 2944(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3136(%rdi), %xmm0
+; SSE-NEXT:    movaps 3072(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3264(%rdi), %xmm0
+; SSE-NEXT:    movaps 3200(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3392(%rdi), %xmm0
+; SSE-NEXT:    movaps 3328(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3520(%rdi), %xmm0
+; SSE-NEXT:    movaps 3456(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3648(%rdi), %xmm0
+; SSE-NEXT:    movaps 3584(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3776(%rdi), %xmm0
+; SSE-NEXT:    movaps 3712(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3904(%rdi), %xmm0
+; SSE-NEXT:    movaps 3840(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 4032(%rdi), %xmm0
+; SSE-NEXT:    movaps 3968(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%rdi), %xmm0
+; SSE-NEXT:    movaps 16(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 208(%rdi), %xmm0
+; SSE-NEXT:    movaps 144(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 336(%rdi), %xmm0
+; SSE-NEXT:    movaps 272(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 464(%rdi), %xmm0
+; SSE-NEXT:    movaps 400(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 592(%rdi), %xmm0
+; SSE-NEXT:    movaps 528(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 720(%rdi), %xmm0
+; SSE-NEXT:    movaps 656(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 848(%rdi), %xmm0
+; SSE-NEXT:    movaps 784(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 976(%rdi), %xmm0
+; SSE-NEXT:    movaps 912(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1104(%rdi), %xmm0
+; SSE-NEXT:    movaps 1040(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1232(%rdi), %xmm0
+; SSE-NEXT:    movaps 1168(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1360(%rdi), %xmm0
+; SSE-NEXT:    movaps 1296(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1488(%rdi), %xmm0
+; SSE-NEXT:    movaps 1424(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1616(%rdi), %xmm0
+; SSE-NEXT:    movaps 1552(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1744(%rdi), %xmm0
+; SSE-NEXT:    movaps 1680(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1872(%rdi), %xmm0
+; SSE-NEXT:    movaps 1808(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2000(%rdi), %xmm0
+; SSE-NEXT:    movaps 1936(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2128(%rdi), %xmm0
+; SSE-NEXT:    movaps 2064(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2256(%rdi), %xmm0
+; SSE-NEXT:    movaps 2192(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2384(%rdi), %xmm0
+; SSE-NEXT:    movaps 2320(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2512(%rdi), %xmm0
+; SSE-NEXT:    movaps 2448(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2640(%rdi), %xmm0
+; SSE-NEXT:    movaps 2576(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2768(%rdi), %xmm0
+; SSE-NEXT:    movaps 2704(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2896(%rdi), %xmm0
+; SSE-NEXT:    movaps 2832(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3024(%rdi), %xmm0
+; SSE-NEXT:    movaps 2960(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3152(%rdi), %xmm0
+; SSE-NEXT:    movaps 3088(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3280(%rdi), %xmm0
+; SSE-NEXT:    movaps 3216(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3408(%rdi), %xmm0
+; SSE-NEXT:    movaps 3344(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3536(%rdi), %xmm0
+; SSE-NEXT:    movaps 3472(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3664(%rdi), %xmm0
+; SSE-NEXT:    movaps 3600(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3792(%rdi), %xmm0
+; SSE-NEXT:    movaps 3728(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3920(%rdi), %xmm0
+; SSE-NEXT:    movaps 3856(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 4048(%rdi), %xmm0
+; SSE-NEXT:    movaps 3984(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 96(%rdi), %xmm0
+; SSE-NEXT:    movaps 32(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 224(%rdi), %xmm0
+; SSE-NEXT:    movaps 160(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 352(%rdi), %xmm0
+; SSE-NEXT:    movaps 288(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 480(%rdi), %xmm0
+; SSE-NEXT:    movaps 416(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 608(%rdi), %xmm0
+; SSE-NEXT:    movaps 544(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 736(%rdi), %xmm0
+; SSE-NEXT:    movaps 672(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 864(%rdi), %xmm0
+; SSE-NEXT:    movaps 800(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 992(%rdi), %xmm0
+; SSE-NEXT:    movaps 928(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1120(%rdi), %xmm0
+; SSE-NEXT:    movaps 1056(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1248(%rdi), %xmm0
+; SSE-NEXT:    movaps 1184(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1376(%rdi), %xmm0
+; SSE-NEXT:    movaps 1312(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1504(%rdi), %xmm0
+; SSE-NEXT:    movaps 1440(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1632(%rdi), %xmm0
+; SSE-NEXT:    movaps 1568(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1760(%rdi), %xmm0
+; SSE-NEXT:    movaps 1696(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1888(%rdi), %xmm0
+; SSE-NEXT:    movaps 1824(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2016(%rdi), %xmm0
+; SSE-NEXT:    movaps 1952(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2144(%rdi), %xmm0
+; SSE-NEXT:    movaps 2080(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2272(%rdi), %xmm0
+; SSE-NEXT:    movaps 2208(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2400(%rdi), %xmm0
+; SSE-NEXT:    movaps 2336(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2528(%rdi), %xmm0
+; SSE-NEXT:    movaps 2464(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2656(%rdi), %xmm0
+; SSE-NEXT:    movaps 2592(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2784(%rdi), %xmm0
+; SSE-NEXT:    movaps 2720(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2912(%rdi), %xmm0
+; SSE-NEXT:    movaps 2848(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3040(%rdi), %xmm0
+; SSE-NEXT:    movaps 2976(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3168(%rdi), %xmm0
+; SSE-NEXT:    movaps 3104(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3296(%rdi), %xmm0
+; SSE-NEXT:    movaps 3232(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3424(%rdi), %xmm0
+; SSE-NEXT:    movaps 3360(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3552(%rdi), %xmm0
+; SSE-NEXT:    movaps 3488(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3680(%rdi), %xmm0
+; SSE-NEXT:    movaps 3616(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3808(%rdi), %xmm0
+; SSE-NEXT:    movaps 3744(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 3936(%rdi), %xmm0
+; SSE-NEXT:    movaps 3872(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 4064(%rdi), %xmm0
+; SSE-NEXT:    movaps 4000(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 112(%rdi), %xmm0
+; SSE-NEXT:    movaps 48(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 240(%rdi), %xmm0
+; SSE-NEXT:    movaps 176(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 368(%rdi), %xmm0
+; SSE-NEXT:    movaps 304(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 496(%rdi), %xmm0
+; SSE-NEXT:    movaps 432(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 624(%rdi), %xmm0
+; SSE-NEXT:    movaps 560(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 752(%rdi), %xmm0
+; SSE-NEXT:    movaps 688(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 880(%rdi), %xmm0
+; SSE-NEXT:    movaps 816(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1008(%rdi), %xmm0
+; SSE-NEXT:    movaps 944(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1136(%rdi), %xmm0
+; SSE-NEXT:    movaps 1072(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1264(%rdi), %xmm0
+; SSE-NEXT:    movaps 1200(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1392(%rdi), %xmm0
+; SSE-NEXT:    movaps 1328(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1520(%rdi), %xmm0
+; SSE-NEXT:    movaps 1456(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1648(%rdi), %xmm0
+; SSE-NEXT:    movaps 1584(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1776(%rdi), %xmm0
+; SSE-NEXT:    movaps 1712(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 1904(%rdi), %xmm0
+; SSE-NEXT:    movaps 1840(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2032(%rdi), %xmm0
+; SSE-NEXT:    movaps 1968(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2160(%rdi), %xmm0
+; SSE-NEXT:    movaps 2096(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2288(%rdi), %xmm0
+; SSE-NEXT:    movaps 2224(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2416(%rdi), %xmm0
+; SSE-NEXT:    movaps 2352(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2544(%rdi), %xmm0
+; SSE-NEXT:    movaps 2480(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2672(%rdi), %xmm0
+; SSE-NEXT:    movaps 2608(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2800(%rdi), %xmm0
+; SSE-NEXT:    movaps 2736(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 2928(%rdi), %xmm0
+; SSE-NEXT:    movaps 2864(%rdi), %xmm15
+; SSE-NEXT:    movaps %xmm15, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm0[1]
+; SSE-NEXT:    movaps 3056(%rdi), %xmm0
+; SSE-NEXT:    movaps 2992(%rdi), %xmm13
+; SSE-NEXT:    movaps %xmm13, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm0[1]
+; SSE-NEXT:    movaps 3184(%rdi), %xmm0
+; SSE-NEXT:    movaps 3120(%rdi), %xmm9
+; SSE-NEXT:    movaps %xmm9, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm0[1]
+; SSE-NEXT:    movaps 3312(%rdi), %xmm0
+; SSE-NEXT:    movaps 3248(%rdi), %xmm8
+; SSE-NEXT:    movaps %xmm8, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
+; SSE-NEXT:    movaps 3440(%rdi), %xmm0
+; SSE-NEXT:    movaps 3376(%rdi), %xmm12
+; SSE-NEXT:    movaps %xmm12, %xmm1
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm0[1]
+; SSE-NEXT:    movaps 3568(%rdi), %xmm0
+; SSE-NEXT:    movaps 3504(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, %xmm14
+; SSE-NEXT:    movlhps {{.*#+}} xmm14 = xmm14[0],xmm0[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm0[1]
+; SSE-NEXT:    movaps 4016(%rdi), %xmm2
+; SSE-NEXT:    movaps 3952(%rdi), %xmm4
+; SSE-NEXT:    movaps 3696(%rdi), %xmm0
+; SSE-NEXT:    movaps 3632(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, %xmm11
+; SSE-NEXT:    movlhps {{.*#+}} xmm11 = xmm11[0],xmm0[0]
+; SSE-NEXT:    movaps 4080(%rdi), %xmm1
+; SSE-NEXT:    movaps 3888(%rdi), %xmm5
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm0[1]
+; SSE-NEXT:    movaps 3824(%rdi), %xmm6
+; SSE-NEXT:    movaps 3760(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm7, 496(%rsi)
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm6[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm6[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, 480(%rsi)
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm4[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm4[1]
+; SSE-NEXT:    movaps %xmm2, %xmm4
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm1[0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm1[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 464(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 448(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 432(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 416(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 400(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 384(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 368(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 352(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 336(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 320(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 304(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 288(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 272(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 256(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 496(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 480(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 464(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 448(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 432(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 416(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 400(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 384(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 368(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 352(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 336(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 320(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 304(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 288(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 272(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 256(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 496(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 480(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 464(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 448(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 432(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 416(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 400(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 384(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 368(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 352(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 336(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 320(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 304(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 288(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 272(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 256(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 496(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 480(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 464(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 448(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 432(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 416(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 400(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 384(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 368(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 352(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 336(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 320(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 304(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 288(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 272(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 256(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 496(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 480(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 464(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 448(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 432(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 416(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 400(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 384(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 368(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 352(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 336(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 320(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 304(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 288(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 272(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 256(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%r9)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 496(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 480(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 464(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 448(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 432(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 416(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 400(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 384(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 368(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 352(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 336(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 320(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 304(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 288(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 272(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 256(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rax)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm4, 496(%rax)
+; SSE-NEXT:    movaps %xmm6, 480(%rax)
+; SSE-NEXT:    movaps %xmm7, 464(%rax)
+; SSE-NEXT:    movaps %xmm11, 448(%rax)
+; SSE-NEXT:    movaps %xmm14, 432(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 416(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 400(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 384(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 368(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 352(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 336(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 320(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 304(%rax)
+; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 288(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 272(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 256(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 240(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 224(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 208(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 192(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 160(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 144(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rax)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm2, 496(%rax)
+; SSE-NEXT:    movaps %xmm5, 480(%rax)
+; SSE-NEXT:    movaps %xmm0, 464(%rax)
+; SSE-NEXT:    movaps %xmm3, 448(%rax)
+; SSE-NEXT:    movaps %xmm10, 432(%rax)
+; SSE-NEXT:    movaps %xmm12, 416(%rax)
+; SSE-NEXT:    movaps %xmm8, 400(%rax)
+; SSE-NEXT:    movaps %xmm9, 384(%rax)
+; SSE-NEXT:    movaps %xmm13, 368(%rax)
+; SSE-NEXT:    movaps %xmm15, 352(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 272(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 256(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 160(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    addq $3720, %rsp # imm = 0xE88
+; SSE-NEXT:    retq
+;
+; AVX1-ONLY-LABEL: load_i64_stride8_vf64:
+; AVX1-ONLY:       # %bb.0:
+; AVX1-ONLY-NEXT:    subq $4968, %rsp # imm = 0x1368
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2240(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 2176(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm6[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vmovaps 2112(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps 2048(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm11[0],xmm10[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm11[1],xmm10[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1024(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm6[0],xmm9[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm6[1],xmm9[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm8[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm8[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1728(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1600(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2752(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2688(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2624(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2560(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3264(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3200(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3136(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3072(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3776(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3712(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3648(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3584(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2496(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2432(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2368(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2304(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3008(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2944(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2880(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2816(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3520(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3456(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3392(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3328(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 4032(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3968(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3904(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3840(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1472(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1280(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1984(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1920(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1856(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1792(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 592(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 528(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1152(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1104(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1040(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1728(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1616(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1552(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2240(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2176(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 2128(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2064(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2752(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2688(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 2640(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2576(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3264(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3200(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 3152(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3088(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3776(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3712(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 3664(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3600(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 4032(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3968(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 3920(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3856(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3520(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3456(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 3408(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3344(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3008(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2944(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 2896(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2832(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2496(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2432(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 2384(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2320(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1984(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1920(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 1872(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vmovaps 1808(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm13[0],xmm14[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1472(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %ymm11
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
+; AVX1-ONLY-NEXT:    vmovaps 1360(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps 1296(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm10[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX1-ONLY-NEXT:    vmovaps 848(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 784(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm5[0],xmm6[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vmovaps 336(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 272(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm9[1],xmm10[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm13[1],xmm14[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1312(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1504(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1632(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1760(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1888(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1824(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2016(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 1952(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2144(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2080(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2272(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2208(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2400(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2336(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2528(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2464(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2656(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2592(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2784(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2720(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2912(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2848(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3040(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 2976(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3168(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3104(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3296(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3232(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3424(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3360(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3552(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3488(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3680(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3616(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3808(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3744(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3936(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 3872(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 4064(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 4000(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 368(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 304(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 624(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 560(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 880(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 816(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm3[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1136(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1072(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1504(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1392(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1328(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm5[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1760(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1696(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1648(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1584(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm6[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2016(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1952(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 1904(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 1840(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm7[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm8[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2272(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2208(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm9 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 2160(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2096(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm10 = xmm8[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2528(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2464(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 2416(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2352(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm11 = xmm9[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2784(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2720(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm11 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 2672(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2608(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm12 = xmm10[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm11[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3040(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2976(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 2928(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 2864(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm11[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3296(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3232(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 3184(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3120(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm12[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm13[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3552(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3488(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 3440(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3376(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm13[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm14[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3808(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3744(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 3696(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3632(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm14[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 4064(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 4000(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovaps 3952(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 3888(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm15[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm3[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm4[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm5[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm6[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm7[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm8[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm9[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm10[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm11[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm12[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm13[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm14[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 496(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 480(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 432(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 416(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 368(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 352(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 288(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 224(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 160(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 48(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 304(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 240(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 176(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 112(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 464(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 448(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 256(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 384(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 320(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 192(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 128(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, (%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 272(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 400(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 336(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 208(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 144(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 80(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 16(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 224(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 240(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 160(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 176(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 112(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 48(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 480(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 496(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 416(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 432(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 352(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 368(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 288(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 304(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 448(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 464(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 384(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 400(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 320(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 336(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 192(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 208(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, (%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 16(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 80(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 128(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 144(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 256(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 272(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 288(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 352(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 416(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 480(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 448(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 384(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 320(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 256(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 480(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 416(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 352(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 288(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 448(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 384(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 320(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 256(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%r8)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%r8)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 496(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 480(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 464(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 448(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 432(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 416(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 400(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 384(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 368(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 352(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 336(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 320(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 304(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 288(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 272(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 256(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 240(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 224(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 208(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 192(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 176(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 160(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 144(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 112(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 96(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 80(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 64(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 48(%r9)
+; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 32(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 16(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, (%r9)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 496(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 480(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 464(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 432(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 416(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 400(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 384(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 368(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 352(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 336(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 320(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 304(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 288(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 272(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 256(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 240(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 208(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 176(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 144(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 112(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 80(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 48(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 16(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, (%rax)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 480(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 416(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 384(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 352(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 320(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 288(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 256(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rax)
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 480(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 416(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 384(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 352(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 320(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm6, 288(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 256(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%rax)
+; AVX1-ONLY-NEXT:    addq $4968, %rsp # imm = 0x1368
+; AVX1-ONLY-NEXT:    vzeroupper
+; AVX1-ONLY-NEXT:    retq
+;
+; AVX2-ONLY-LABEL: load_i64_stride8_vf64:
+; AVX2-ONLY:       # %bb.0:
+; AVX2-ONLY-NEXT:    subq $5560, %rsp # imm = 0x15B8
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 704(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps 640(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vmovaps 1152(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2240(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vmovaps 2176(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm8[0],xmm7[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 576(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vmovaps 1088(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vmovaps 2112(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vmovaps 2048(%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm8[1],xmm7[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm11[0],xmm10[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm11[1],xmm10[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1024(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm7[0],xmm9[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm7[1],xmm9[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 512(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm6[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm6[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1728(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1664(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1600(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1536(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2752(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2688(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2624(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2560(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3264(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3200(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3136(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3072(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3776(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3712(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3648(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3584(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2496(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2432(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2368(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2304(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3008(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2944(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2880(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2816(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3520(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3456(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3392(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3328(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 4032(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3968(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3904(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3840(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 960(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 896(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 832(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 768(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1472(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1280(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1984(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1920(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1856(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1792(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 576(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 512(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 704(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 640(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1088(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1024(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1216(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1152(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1600(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1536(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1728(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1664(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2112(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2048(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2240(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2176(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2624(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2560(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2752(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2688(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3136(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3072(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3264(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3200(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3648(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3584(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3776(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3712(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3904(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3840(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 4032(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3968(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3392(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3328(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3520(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3456(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2880(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2816(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3008(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2944(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2368(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2304(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2496(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2432(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1856(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1792(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1984(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vmovaps 1920(%rdi), %ymm10
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm11[0],ymm10[2],ymm11[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1344(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovaps 1280(%rdi), %ymm14
+; AVX2-ONLY-NEXT:    vmovaps 1472(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovaps 1408(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm14[0],ymm9[0],ymm14[2],ymm9[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 832(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovaps 768(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vmovaps 960(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovaps 896(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm13[0],ymm6[0],ymm13[2],ymm6[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %ymm12
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm12[0],ymm3[0],ymm12[2],ymm3[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm12[1],ymm3[1],ymm12[3],ymm3[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm13[1],ymm6[1],ymm13[3],ymm6[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm14[1],ymm9[1],ymm14[3],ymm9[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm11[1],ymm10[3],ymm11[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 608(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 544(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 736(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 672(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 864(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 800(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 992(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 928(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1120(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1248(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1312(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1504(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1440(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1632(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1760(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1696(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1888(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1824(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2016(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 1952(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2144(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2080(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2272(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2208(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2400(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2336(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2528(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2464(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2656(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2592(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2784(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2720(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2912(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2848(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3040(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 2976(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3168(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3104(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3296(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3232(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3424(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3360(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3552(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3488(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3680(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3616(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3808(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3744(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3936(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 3872(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 4064(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 4000(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 608(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 544(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 736(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 672(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 864(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 800(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 992(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 928(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1120(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1056(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1184(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1376(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1312(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1504(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1632(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1568(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1760(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1696(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1888(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1824(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2016(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 1952(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2144(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2080(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2272(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2208(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2400(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2336(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2528(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2464(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2656(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2592(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2784(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2720(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2912(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2848(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3040(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 2976(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3168(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3104(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3296(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vmovaps 3232(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm9[0],ymm11[0],ymm9[2],ymm11[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3424(%rdi), %ymm10
+; AVX2-ONLY-NEXT:    vmovaps 3360(%rdi), %ymm14
+; AVX2-ONLY-NEXT:    vmovaps 3552(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovaps 3488(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm14[0],ymm10[0],ymm14[2],ymm10[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3680(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovaps 3616(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vmovaps 3808(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovaps 3744(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm5[0],ymm3[2],ymm5[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm13[0],ymm6[0],ymm13[2],ymm6[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 3936(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovaps 3872(%rdi), %ymm12
+; AVX2-ONLY-NEXT:    vmovaps 4064(%rdi), %ymm2
+; AVX2-ONLY-NEXT:    vmovaps 4000(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm12[0],ymm4[0],ymm12[2],ymm4[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm9[1],ymm11[1],ymm9[3],ymm11[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm9 = ymm9[1],mem[1],ymm9[3],mem[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm9[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm7 = ymm14[1],ymm10[1],ymm14[3],ymm10[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm7[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm5[1],ymm3[3],ymm5[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm13[1],ymm6[1],ymm13[3],ymm6[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm12[1],ymm4[1],ymm12[3],ymm4[3]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 496(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 480(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 432(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 416(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 368(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 352(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 32(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 288(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 224(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 160(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 96(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 48(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 304(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 240(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 176(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 112(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 464(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 448(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 256(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 384(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 320(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 192(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 128(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 64(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, (%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 272(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 400(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 336(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 208(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 144(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 80(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 16(%rsi)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 224(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 240(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 160(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 176(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 96(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 112(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 32(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 48(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 480(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 496(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 416(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 432(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 352(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 368(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 288(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 304(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 448(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 464(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 384(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 400(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 320(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 336(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 192(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 208(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, (%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 16(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 64(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 80(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 128(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 144(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 256(%rdx)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 272(%rdx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 416(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 480(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rcx)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 480(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 416(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%r8)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 496(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 480(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 464(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 448(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 432(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 416(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 400(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 384(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 368(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 352(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 336(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 320(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 304(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 288(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 272(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 256(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 240(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 224(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 208(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 192(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 176(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 160(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 144(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 128(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 112(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 96(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 80(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 64(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 48(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 32(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 16(%r9)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, (%r9)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 496(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 480(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 464(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 448(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 432(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 416(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 400(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 384(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 368(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 352(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 336(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 320(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 304(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 288(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 272(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 256(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 240(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 208(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 176(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 160(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 144(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 112(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 80(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 48(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 16(%rax)
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, (%rax)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 480(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 448(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 416(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rax)
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 480(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm3, 448(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 416(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm9, 384(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm15, 352(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 320(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 288(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 256(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 96(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, (%rax)
+; AVX2-ONLY-NEXT:    addq $5560, %rsp # imm = 0x15B8
+; AVX2-ONLY-NEXT:    vzeroupper
+; AVX2-ONLY-NEXT:    retq
+;
+; AVX512F-LABEL: load_i64_stride8_vf64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    subq $6728, %rsp # imm = 0x1A48
+; AVX512F-NEXT:    vmovdqa64 3392(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 3328(%rdi), %zmm13
+; AVX512F-NEXT:    vmovdqa64 3520(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 3456(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1856(%rdi), %zmm11
+; AVX512F-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1984(%rdi), %zmm15
+; AVX512F-NEXT:    vmovdqu64 %zmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 832(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 768(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 960(%rdi), %zmm24
+; AVX512F-NEXT:    vmovdqa64 896(%rdi), %zmm10
+; AVX512F-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 320(%rdi), %zmm12
+; AVX512F-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 256(%rdi), %zmm5
+; AVX512F-NEXT:    vmovdqa64 448(%rdi), %zmm8
+; AVX512F-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 384(%rdi), %zmm14
+; AVX512F-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    movb $-64, %al
+; AVX512F-NEXT:    kmovw %eax, %k1
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [2,10,2,10,2,10,2,10]
+; AVX512F-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqa64 3264(%rdi), %ymm21
+; AVX512F-NEXT:    vmovdqa 3200(%rdi), %ymm0
+; AVX512F-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm21[0],ymm0[2],ymm21[2]
+; AVX512F-NEXT:    vmovdqa 3136(%rdi), %ymm4
+; AVX512F-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vmovdqa 3072(%rdi), %ymm3
+; AVX512F-NEXT:    vmovdqu %ymm3, (%rsp) # 32-byte Spill
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm24, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm6, %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqa 704(%rdi), %ymm3
+; AVX512F-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vmovdqa 640(%rdi), %ymm0
+; AVX512F-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %ymm25
+; AVX512F-NEXT:    vmovdqa64 512(%rdi), %ymm23
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm23[0],ymm25[0],ymm23[2],ymm25[2]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm14, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm12, %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqa 192(%rdi), %ymm0
+; AVX512F-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vmovdqa64 128(%rdi), %ymm31
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm31[0],ymm0[0],ymm31[2],ymm0[2]
+; AVX512F-NEXT:    vmovdqa64 64(%rdi), %ymm20
+; AVX512F-NEXT:    vmovdqa64 (%rdi), %ymm19
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm19[0],ymm20[0],ymm19[2],ymm20[2]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512F-NEXT:    vmovdqa64 1920(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqa64 1792(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm11, %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqa 1728(%rdi), %ymm3
+; AVX512F-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vmovdqa 1664(%rdi), %ymm0
+; AVX512F-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; AVX512F-NEXT:    vmovdqa64 1600(%rdi), %ymm18
+; AVX512F-NEXT:    vmovdqa64 1536(%rdi), %ymm26
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm26[0],ymm18[0],ymm26[2],ymm18[2]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1472(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1408(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqa64 1344(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1280(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqa64 1216(%rdi), %ymm28
+; AVX512F-NEXT:    vmovdqa64 1152(%rdi), %ymm29
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm29[0],ymm28[0],ymm29[2],ymm28[2]
+; AVX512F-NEXT:    vmovdqa64 1088(%rdi), %ymm30
+; AVX512F-NEXT:    vmovdqa64 1024(%rdi), %ymm27
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm27[0],ymm30[0],ymm27[2],ymm30[2]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 3008(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2944(%rdi), %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqa64 2880(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2816(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqa 2752(%rdi), %ymm0
+; AVX512F-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vmovdqa 2688(%rdi), %ymm11
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm11[0],ymm0[0],ymm11[2],ymm0[2]
+; AVX512F-NEXT:    vmovdqa64 2624(%rdi), %ymm16
+; AVX512F-NEXT:    vmovdqa 2560(%rdi), %ymm9
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm9[0],ymm16[0],ymm9[2],ymm16[2]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2496(%rdi), %zmm10
+; AVX512F-NEXT:    vmovdqa64 2432(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2368(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2304(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm3, %zmm2, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovdqa 2240(%rdi), %ymm0
+; AVX512F-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vmovdqa 2176(%rdi), %ymm3
+; AVX512F-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX512F-NEXT:    vmovdqa 2112(%rdi), %ymm3
+; AVX512F-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-NEXT:    vmovdqa 2048(%rdi), %ymm8
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm8[0],ymm3[0],ymm8[2],ymm3[2]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 4032(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 3968(%rdi), %zmm6
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqa64 3904(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqa64 3840(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm1, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqa64 3776(%rdi), %ymm22
+; AVX512F-NEXT:    vmovdqa64 3712(%rdi), %ymm17
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm15 = ymm17[0],ymm22[0],ymm17[2],ymm22[2]
+; AVX512F-NEXT:    vmovdqa 3648(%rdi), %ymm12
+; AVX512F-NEXT:    vmovdqa 3584(%rdi), %ymm0
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} ymm14 = ymm0[0],ymm12[0],ymm0[2],ymm12[2]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm14 = ymm14[2,3],ymm15[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm14, %zmm2, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [3,11,3,11,3,11,3,11]
+; AVX512F-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm14
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm2, %zmm14
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm15
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm15 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm14, %zmm15 {%k1}
+; AVX512F-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm14 = ymm13[1],ymm21[1],ymm13[3],ymm21[3]
+; AVX512F-NEXT:    vmovdqu (%rsp), %ymm13 # 32-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm13[2,3],ymm14[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm13, %zmm15, %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm2, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm14
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm14 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm14 {%k1}
+; AVX512F-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm15 = ymm23[1],ymm25[1],ymm23[3],ymm25[3]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm15[2,3],ymm13[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm13, %zmm14, %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm25, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm2, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm23 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm23, %zmm2, %zmm5
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm5 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm31, %ymm13 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm13 = ymm31[1],mem[1],ymm31[3],mem[3]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm15 = ymm19[1],ymm20[1],ymm19[3],ymm20[3]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm15[2,3],ymm13[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm13, %zmm5, %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm20, %zmm2, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm2, %zmm14
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm14 {%k1}
+; AVX512F-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm13 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm15 = ymm26[1],ymm18[1],ymm26[3],ymm18[3]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm15[2,3],ymm13[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm13, %zmm14, %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm2, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm14 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm14 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm13 = ymm29[1],ymm28[1],ymm29[3],ymm28[3]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm15 = ymm27[1],ymm30[1],ymm27[3],ymm30[3]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm15[2,3],ymm13[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm13, %zmm14, %zmm13
+; AVX512F-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm2, %zmm13
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm30, %zmm14
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm29 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm2, %zmm14
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm14 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm11 = ymm11[1],mem[1],ymm11[3],mem[3]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm9 = ymm9[1],ymm16[1],ymm9[3],ymm16[3]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm9 = ymm9[2,3],ymm11[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm9, %zmm14, %zmm9
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm2, %zmm7
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm16, %zmm10
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm10 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm7, %zmm10 {%k1}
+; AVX512F-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm7 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm7 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm5 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm5 = ymm8[1],mem[1],ymm8[3],mem[3]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm5 = ymm5[2,3],ymm7[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm5, %zmm10, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm2, %zmm6
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2q %zmm3, %zmm8, %zmm2
+; AVX512F-NEXT:    vmovdqa64 %zmm6, %zmm2 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm3 = ymm17[1],ymm22[1],ymm17[3],ymm22[3]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm12[1],ymm0[3],ymm12[3]
+; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm3[2,3]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm2, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [4,12,4,12,4,12,4,12]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm13
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm11
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm0, %zmm1
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm2 = zmm22[0],zmm9[0],zmm22[2],zmm9[2],zmm22[4],zmm9[4],zmm22[6],zmm9[6]
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512F-NEXT:    vmovdqa64 3136(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 3072(%rdi), %zmm15
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm4, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqa64 3264(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 3200(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm6 = <u,u,4,12>
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm6, %zmm4
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm3, %zmm2, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm21[0],zmm17[0],zmm21[2],zmm17[2],zmm21[4],zmm17[4],zmm21[6],zmm17[6]
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 512(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 704(%rdi), %zmm21
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm6, %zmm4
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm25, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm12[0],zmm23[0],zmm12[2],zmm23[2],zmm12[4],zmm23[4],zmm12[6],zmm23[6]
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqa64 64(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 (%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 192(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 128(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm6, %zmm4
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm20, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm1[0],zmm19[0],zmm1[2],zmm19[2],zmm1[4],zmm19[4],zmm1[6],zmm19[6]
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqa64 1600(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1536(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 1728(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1664(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm6, %zmm4
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm1[0],zmm10[0],zmm1[2],zmm10[2],zmm1[4],zmm10[4],zmm1[6],zmm10[6]
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqa64 1088(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1024(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 1216(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 1152(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm6, %zmm4
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm27, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm0, %zmm2
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm30[0],zmm29[0],zmm30[2],zmm29[2],zmm30[4],zmm29[4],zmm30[6],zmm29[6]
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqa64 2624(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2560(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 2752(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2688(%rdi), %zmm4
+; AVX512F-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm6, %zmm4
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2112(%rdi), %zmm7
+; AVX512F-NEXT:    vmovdqa64 2048(%rdi), %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2240(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 2176(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm6, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm26, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm4 = zmm16[0],zmm30[0],zmm16[2],zmm30[2],zmm16[4],zmm30[4],zmm16[6],zmm30[6]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa64 3648(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 3584(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm3, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqa64 3776(%rdi), %zmm3
+; AVX512F-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 3712(%rdi), %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm3, %zmm1, %zmm6
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
+; AVX512F-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm1 # 64-byte Folded Reload
+; AVX512F-NEXT:    # zmm1 = zmm8[0],mem[0],zmm8[2],mem[2],zmm8[4],mem[4],zmm8[6],mem[6]
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [5,13,5,13,5,13,5,13]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm25, %zmm0, %zmm15
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,5,13>
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm15[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vpermt2q %zmm11, %zmm0, %zmm13
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm4 = zmm22[1],zmm9[1],zmm22[3],zmm9[3],zmm22[5],zmm9[5],zmm22[7],zmm9[7]
+; AVX512F-NEXT:    vmovdqa64 %zmm13, %zmm4 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm20, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 %zmm21, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm23 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm23, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm18, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm17, %zmm19
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm4 = zmm17[1],zmm19[1],zmm17[3],zmm19[3],zmm17[5],zmm19[5],zmm17[7],zmm19[7]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm14, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm16, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm3 # 64-byte Folded Reload
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm12, %zmm4 # 64-byte Folded Reload
+; AVX512F-NEXT:    # zmm4 = zmm12[1],mem[1],zmm12[3],mem[3],zmm12[5],mem[5],zmm12[7],mem[7]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm13, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm3 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm4 # 64-byte Folded Reload
+; AVX512F-NEXT:    # zmm4 = zmm4[1],mem[1],zmm4[3],mem[3],zmm4[5],mem[5],zmm4[7],mem[7]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm11, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm29 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm3 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm27
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm4 = zmm4[1],zmm10[1],zmm4[3],zmm10[3],zmm4[5],zmm10[5],zmm4[7],zmm10[7]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm3 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm4 # 64-byte Folded Reload
+; AVX512F-NEXT:    # zmm4 = zmm4[1],mem[1],zmm4[3],mem[3],zmm4[5],mem[5],zmm4[7],mem[7]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm26, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm5, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm4 = zmm5[1],zmm30[1],zmm5[3],zmm30[3],zmm5[5],zmm30[5],zmm5[7],zmm30[7]
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2q %zmm31, %zmm6, %zmm0
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2q %zmm30, %zmm3, %zmm1
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm1 # 64-byte Folded Reload
+; AVX512F-NEXT:    # zmm1 = zmm4[1],mem[1],zmm4[3],mem[3],zmm4[5],mem[5],zmm4[7],mem[7]
+; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1 {%k1}
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [6,14,6,14,6,14,6,14]
+; AVX512F-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm25, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,6,14>
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm24, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm3
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm3 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm22, %zmm3 {%k1} # 64-byte Folded Reload
+; AVX512F-NEXT:    # zmm3 {%k1} = zmm22[0],mem[0],zmm22[2],mem[2],zmm22[4],mem[4],zmm22[6],mem[6]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm20, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm21, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm17, %zmm3
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm0, %zmm3
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 {%k1} = zmm23[0],zmm18[0],zmm23[2],zmm18[2],zmm23[4],zmm18[4],zmm23[6],zmm18[6]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm16, %zmm0, %zmm14
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm14[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm18
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 {%k1} = zmm20[0],zmm14[0],zmm20[2],zmm14[2],zmm20[4],zmm14[4],zmm20[6],zmm14[6]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm13, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 {%k1} = zmm26[0],zmm25[0],zmm26[2],zmm25[2],zmm26[4],zmm25[4],zmm26[6],zmm25[6]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm11, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm27, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm29 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 {%k1} = zmm24[0],zmm29[0],zmm24[2],zmm29[2],zmm24[4],zmm29[4],zmm24[6],zmm29[6]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm1, %zmm3
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm3 {%k1} = zmm22[0],zmm19[0],zmm22[2],zmm19[2],zmm22[4],zmm19[4],zmm22[6],zmm19[6]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 # 64-byte Folded Reload
+; AVX512F-NEXT:    vpermt2q %zmm7, %zmm1, %zmm8
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm0, %zmm5
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm5 {%k1} = zmm21[0],zmm17[0],zmm21[2],zmm17[2],zmm21[4],zmm17[4],zmm21[6],zmm17[6]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm4 # 64-byte Folded Reload
+; AVX512F-NEXT:    vpermi2q %zmm31, %zmm6, %zmm0
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2q %zmm30, %zmm3, %zmm1
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm23 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} zmm4 {%k1} = zmm1[0],zmm23[0],zmm1[2],zmm23[2],zmm1[4],zmm23[4],zmm1[6],zmm23[6]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm0, %zmm4, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm30
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm5
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm16 = [7,15,7,15,7,15,7,15]
+; AVX512F-NEXT:    # zmm16 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm16, %zmm0
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm0 {%k1} = zmm4[1],zmm1[1],zmm4[3],zmm1[3],zmm4[5],zmm1[5],zmm4[7],zmm1[7]
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm6
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm3 = [0,8,0,8,0,8,0,8]
+; AVX512F-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vbroadcasti32x4 {{.*#+}} zmm13 = [1,9,1,9,1,9,1,9]
+; AVX512F-NEXT:    # zmm13 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm13, %zmm6
+; AVX512F-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm3, %zmm30
+; AVX512F-NEXT:    vpermt2q %zmm2, %zmm13, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm18, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm16, %zmm1
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm1 {%k1} = zmm20[1],zmm14[1],zmm20[3],zmm14[3],zmm20[5],zmm14[5],zmm20[7],zmm14[7]
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm3, %zmm20
+; AVX512F-NEXT:    vpermt2q %zmm14, %zmm13, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm3, %zmm18
+; AVX512F-NEXT:    vpermt2q %zmm28, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm28
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm16, %zmm12
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm12 {%k1} = zmm26[1],zmm25[1],zmm26[3],zmm25[3],zmm26[5],zmm25[5],zmm26[7],zmm25[7]
+; AVX512F-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm26, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm25, %zmm3, %zmm26
+; AVX512F-NEXT:    vmovdqu64 %zmm26, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm25, %zmm13, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm3, %zmm28
+; AVX512F-NEXT:    vpermt2q %zmm15, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm25
+; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm27, %zmm16, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm7
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm1 {%k1} = zmm24[1],zmm29[1],zmm24[3],zmm29[3],zmm24[5],zmm29[5],zmm24[7],zmm29[7]
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm3, %zmm7
+; AVX512F-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm29, %zmm13, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm27, %zmm3, %zmm25
+; AVX512F-NEXT:    vpermt2q %zmm27, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm27
+; AVX512F-NEXT:    vmovdqa64 %zmm9, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm16, %zmm9
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm9 {%k1} = zmm22[1],zmm19[1],zmm22[3],zmm19[3],zmm22[5],zmm19[5],zmm22[7],zmm19[7]
+; AVX512F-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm22, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm3, %zmm22
+; AVX512F-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm19, %zmm13, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm3, %zmm27
+; AVX512F-NEXT:    vpermt2q %zmm10, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm29
+; AVX512F-NEXT:    vmovdqa64 %zmm31, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm16, %zmm31
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm31 {%k1} = zmm21[1],zmm17[1],zmm21[3],zmm17[3],zmm21[5],zmm17[5],zmm21[7],zmm17[7]
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm19
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm3, %zmm19
+; AVX512F-NEXT:    vpermt2q %zmm17, %zmm13, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm3, %zmm29
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm22
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm3, %zmm22
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm23, %zmm3, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512F-NEXT:    vpermi2q %zmm26, %zmm15, %zmm3
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm8, %zmm13, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm17
+; AVX512F-NEXT:    vmovdqa64 %zmm21, %zmm0
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm13, %zmm17
+; AVX512F-NEXT:    vmovdqa64 %zmm24, %zmm1
+; AVX512F-NEXT:    vpermt2q %zmm23, %zmm13, %zmm1
+; AVX512F-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpermi2q %zmm26, %zmm15, %zmm13
+; AVX512F-NEXT:    vpermt2q %zmm26, %zmm16, %zmm15
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm15 {%k1} = zmm24[1],zmm23[1],zmm24[3],zmm23[3],zmm24[5],zmm23[5],zmm24[7],zmm23[7]
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm21
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm15
+; AVX512F-NEXT:    vpermt2q %zmm9, %zmm16, %zmm15
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} zmm15 {%k1} = zmm10[1],zmm8[1],zmm10[3],zmm8[3],zmm10[5],zmm8[5],zmm10[7],zmm8[7]
+; AVX512F-NEXT:    vmovdqa64 %zmm15, %zmm23
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm12 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm5 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm6 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm7 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm8 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm10 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm0 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm14 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} ymm16 = <u,u,7,15>
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm15 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm9 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm0 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm1 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm2 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm4 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm11 # 64-byte Folded Reload
+; AVX512F-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512F-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm11 # 64-byte Folded Reload
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm15 = ymm12[0,1,2,3],ymm15[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm15, %zmm12, %zmm12
+; AVX512F-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm15 = ymm5[0,1,2,3],ymm9[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm15, %zmm5, %zmm5
+; AVX512F-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm15 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm15, %zmm0, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm15 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm15, %zmm0, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm15 = ymm8[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm15, %zmm0, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm15 = ymm10[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm15, %zmm31, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vpblendd $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX512F-NEXT:    # ymm15 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm15, %zmm21, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vpblendd {{.*#+}} ymm15 = ymm14[0,1,2,3],ymm11[4,5,6,7]
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm15, %zmm23, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm20, %zmm18 {%k1}
+; AVX512F-NEXT:    vmovdqa 192(%rdi), %xmm0
+; AVX512F-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-NEXT:    vmovdqa 128(%rdi), %xmm1
+; AVX512F-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512F-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-NEXT:    vmovdqa 64(%rdi), %xmm1
+; AVX512F-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm16 = xmm0[0],xmm1[0]
+; AVX512F-NEXT:    vinserti32x4 $1, %xmm2, %ymm16, %ymm2
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm18, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm30 {%k1}
+; AVX512F-NEXT:    vmovdqa 704(%rdi), %xmm8
+; AVX512F-NEXT:    vmovdqa64 640(%rdi), %xmm20
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm20[0],xmm8[0]
+; AVX512F-NEXT:    vmovdqa64 576(%rdi), %xmm18
+; AVX512F-NEXT:    vmovdqa64 512(%rdi), %xmm16
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm31 = xmm16[0],xmm18[0]
+; AVX512F-NEXT:    vinserti32x4 $1, %xmm2, %ymm31, %ymm2
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm30, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm25 {%k1}
+; AVX512F-NEXT:    vmovdqa 1216(%rdi), %xmm0
+; AVX512F-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-NEXT:    vmovdqa 1152(%rdi), %xmm1
+; AVX512F-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm31 = xmm1[0],xmm0[0]
+; AVX512F-NEXT:    vmovdqa64 1088(%rdi), %xmm23
+; AVX512F-NEXT:    vmovdqa 1024(%rdi), %xmm14
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm30 = xmm14[0],xmm23[0]
+; AVX512F-NEXT:    vinserti32x4 $1, %xmm31, %ymm30, %ymm30
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm30, %zmm25, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm28 {%k1}
+; AVX512F-NEXT:    vmovdqa64 1728(%rdi), %xmm30
+; AVX512F-NEXT:    vmovdqa64 1664(%rdi), %xmm31
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm31[0],xmm30[0]
+; AVX512F-NEXT:    vmovdqa 1600(%rdi), %xmm12
+; AVX512F-NEXT:    vmovdqa 1536(%rdi), %xmm11
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm25 = xmm11[0],xmm12[0]
+; AVX512F-NEXT:    vinserti32x4 $1, %xmm2, %ymm25, %ymm2
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm28, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqa64 %zmm19, %zmm29 {%k1}
+; AVX512F-NEXT:    vmovdqa64 2240(%rdi), %xmm19
+; AVX512F-NEXT:    vmovdqa64 2176(%rdi), %xmm21
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm25 = xmm21[0],xmm19[0]
+; AVX512F-NEXT:    vmovdqa 2112(%rdi), %xmm15
+; AVX512F-NEXT:    vmovdqa 2048(%rdi), %xmm1
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm28 = xmm1[0],xmm15[0]
+; AVX512F-NEXT:    vinserti32x4 $1, %xmm25, %ymm28, %ymm25
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm25, %zmm29, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm27 {%k1}
+; AVX512F-NEXT:    vmovdqa64 2752(%rdi), %xmm28
+; AVX512F-NEXT:    vmovdqa64 2688(%rdi), %xmm29
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm29[0],xmm28[0]
+; AVX512F-NEXT:    vmovdqa 2624(%rdi), %xmm10
+; AVX512F-NEXT:    vmovdqa 2560(%rdi), %xmm9
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm25 = xmm9[0],xmm10[0]
+; AVX512F-NEXT:    vinserti32x4 $1, %xmm2, %ymm25, %ymm2
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm27, %zmm0
+; AVX512F-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm22 {%k1}
+; AVX512F-NEXT:    vmovdqa64 3264(%rdi), %xmm25
+; AVX512F-NEXT:    vmovdqa64 3200(%rdi), %xmm27
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm4 = xmm27[0],xmm25[0]
+; AVX512F-NEXT:    vmovdqa 3136(%rdi), %xmm0
+; AVX512F-NEXT:    vmovdqa 3072(%rdi), %xmm2
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm26 = xmm2[0],xmm0[0]
+; AVX512F-NEXT:    vinserti32x4 $1, %xmm4, %ymm26, %ymm4
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm4, %zmm22, %zmm22
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm3 {%k1}
+; AVX512F-NEXT:    vmovdqa 3776(%rdi), %xmm7
+; AVX512F-NEXT:    vmovdqa64 3712(%rdi), %xmm26
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm4 = xmm26[0],xmm7[0]
+; AVX512F-NEXT:    vmovdqa 3648(%rdi), %xmm6
+; AVX512F-NEXT:    vmovdqa 3584(%rdi), %xmm5
+; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm24 = xmm5[0],xmm6[0]
+; AVX512F-NEXT:    vinserti32x4 $1, %xmm4, %ymm24, %ymm4
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm4, %zmm3, %zmm24
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm17 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm4 = xmm27[1],xmm25[1]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; AVX512F-NEXT:    vinserti128 $1, %xmm4, %ymm2, %ymm2
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm2, %zmm17, %zmm2
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm3, %zmm0 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm3 = xmm20[1],xmm8[1]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm4 = xmm16[1],xmm18[1]
+; AVX512F-NEXT:    vinserti128 $1, %xmm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm3, %zmm0, %zmm3
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm0 {%k1}
+; AVX512F-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX512F-NEXT:    # xmm4 = xmm4[1],mem[1]
+; AVX512F-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX512F-NEXT:    # xmm8 = xmm8[1],mem[1]
+; AVX512F-NEXT:    vinserti128 $1, %xmm4, %ymm8, %ymm4
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm4, %zmm0, %zmm4
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm8, %zmm0 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm8 = xmm31[1],xmm30[1]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm11 = xmm11[1],xmm12[1]
+; AVX512F-NEXT:    vinserti128 $1, %xmm8, %ymm11, %ymm8
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm8, %zmm0, %zmm8
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm11, %zmm0 {%k1}
+; AVX512F-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX512F-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
+; AVX512F-NEXT:    # xmm11 = xmm11[1],mem[1]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm12 = xmm14[1],xmm23[1]
+; AVX512F-NEXT:    vinserti128 $1, %xmm11, %ymm12, %ymm11
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm11, %zmm0, %zmm11
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm12, %zmm0 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm12 = xmm29[1],xmm28[1]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm9 = xmm9[1],xmm10[1]
+; AVX512F-NEXT:    vinserti128 $1, %xmm12, %ymm9, %ymm9
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm9, %zmm0, %zmm9
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm10, %zmm0 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm10 = xmm21[1],xmm19[1]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm12 = xmm1[1],xmm15[1]
+; AVX512F-NEXT:    vinserti128 $1, %xmm10, %ymm12, %ymm10
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm10, %zmm0, %zmm10
+; AVX512F-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovdqa64 %zmm0, %zmm13 {%k1}
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm7 = xmm26[1],xmm7[1]
+; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm5 = xmm5[1],xmm6[1]
+; AVX512F-NEXT:    vinserti128 $1, %xmm7, %ymm5, %ymm5
+; AVX512F-NEXT:    vinserti64x4 $0, %ymm5, %zmm13, %zmm1
+; AVX512F-NEXT:    vmovdqa64 %zmm24, 448(%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm22, 384(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 320(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 256(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%rsi)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%rsi)
+; AVX512F-NEXT:    vmovdqa64 %zmm1, 448(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm10, 256(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm9, 320(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm11, 128(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm8, 192(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm4, (%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm3, 64(%rdx)
+; AVX512F-NEXT:    vmovdqa64 %zmm2, 384(%rdx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 448(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 256(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 320(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 384(%rcx)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 448(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 256(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 320(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 384(%r8)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 448(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 256(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 320(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%r9)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 384(%r9)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 448(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 256(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 320(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 384(%rax)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 448(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 256(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 320(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 384(%rax)
+; AVX512F-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-NEXT:    vmovups (%rsp), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 384(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 448(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 256(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 320(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 128(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 192(%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, (%rax)
+; AVX512F-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-NEXT:    vmovaps %zmm0, 64(%rax)
+; AVX512F-NEXT:    addq $6728, %rsp # imm = 0x1A48
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: load_i64_stride8_vf64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    subq $6728, %rsp # imm = 0x1A48
+; AVX512BW-NEXT:    vmovdqa64 3392(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 3328(%rdi), %zmm13
+; AVX512BW-NEXT:    vmovdqa64 3520(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 3456(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1856(%rdi), %zmm11
+; AVX512BW-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1984(%rdi), %zmm15
+; AVX512BW-NEXT:    vmovdqu64 %zmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm24
+; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm10
+; AVX512BW-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm12
+; AVX512BW-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm5
+; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm8
+; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm14
+; AVX512BW-NEXT:    vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    movb $-64, %al
+; AVX512BW-NEXT:    kmovd %eax, %k1
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [2,10,2,10,2,10,2,10]
+; AVX512BW-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 3264(%rdi), %ymm21
+; AVX512BW-NEXT:    vmovdqa 3200(%rdi), %ymm0
+; AVX512BW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm21[0],ymm0[2],ymm21[2]
+; AVX512BW-NEXT:    vmovdqa 3136(%rdi), %ymm4
+; AVX512BW-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vmovdqa 3072(%rdi), %ymm3
+; AVX512BW-NEXT:    vmovdqu %ymm3, (%rsp) # 32-byte Spill
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm24, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa 704(%rdi), %ymm3
+; AVX512BW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vmovdqa 640(%rdi), %ymm0
+; AVX512BW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %ymm25
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %ymm23
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm23[0],ymm25[0],ymm23[2],ymm25[2]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa 192(%rdi), %ymm0
+; AVX512BW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %ymm31
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm31[0],ymm0[0],ymm31[2],ymm0[2]
+; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %ymm20
+; AVX512BW-NEXT:    vmovdqa64 (%rdi), %ymm19
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm19[0],ymm20[0],ymm19[2],ymm20[2]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512BW-NEXT:    vmovdqa64 1920(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 1792(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa 1728(%rdi), %ymm3
+; AVX512BW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vmovdqa 1664(%rdi), %ymm0
+; AVX512BW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; AVX512BW-NEXT:    vmovdqa64 1600(%rdi), %ymm18
+; AVX512BW-NEXT:    vmovdqa64 1536(%rdi), %ymm26
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm26[0],ymm18[0],ymm26[2],ymm18[2]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1472(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1408(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 1344(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1280(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %ymm28
+; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %ymm29
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm29[0],ymm28[0],ymm29[2],ymm28[2]
+; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %ymm30
+; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %ymm27
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm27[0],ymm30[0],ymm27[2],ymm30[2]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 3008(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2944(%rdi), %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 2880(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2816(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa 2752(%rdi), %ymm0
+; AVX512BW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vmovdqa 2688(%rdi), %ymm11
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm11[0],ymm0[0],ymm11[2],ymm0[2]
+; AVX512BW-NEXT:    vmovdqa64 2624(%rdi), %ymm16
+; AVX512BW-NEXT:    vmovdqa 2560(%rdi), %ymm9
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm9[0],ymm16[0],ymm9[2],ymm16[2]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2496(%rdi), %zmm10
+; AVX512BW-NEXT:    vmovdqa64 2432(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2368(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2304(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; AVX512BW-NEXT:    vmovdqa 2240(%rdi), %ymm0
+; AVX512BW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vmovdqa 2176(%rdi), %ymm3
+; AVX512BW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX512BW-NEXT:    vmovdqa 2112(%rdi), %ymm3
+; AVX512BW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512BW-NEXT:    vmovdqa 2048(%rdi), %ymm8
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm8[0],ymm3[0],ymm8[2],ymm3[2]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 4032(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 3968(%rdi), %zmm6
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 3904(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqa64 3840(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 3776(%rdi), %ymm22
+; AVX512BW-NEXT:    vmovdqa64 3712(%rdi), %ymm17
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm15 = ymm17[0],ymm22[0],ymm17[2],ymm22[2]
+; AVX512BW-NEXT:    vmovdqa 3648(%rdi), %ymm12
+; AVX512BW-NEXT:    vmovdqa 3584(%rdi), %ymm0
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm14 = ymm0[0],ymm12[0],ymm0[2],ymm12[2]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm14 = ymm14[2,3],ymm15[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm14, %zmm2, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [3,11,3,11,3,11,3,11]
+; AVX512BW-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm2, %zmm14
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm15
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm15 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, %zmm15 {%k1}
+; AVX512BW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm14 = ymm13[1],ymm21[1],ymm13[3],ymm21[3]
+; AVX512BW-NEXT:    vmovdqu (%rsp), %ymm13 # 32-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm13[2,3],ymm14[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm13, %zmm15, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm2, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm14
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm14 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm14 {%k1}
+; AVX512BW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm15 = ymm23[1],ymm25[1],ymm23[3],ymm25[3]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm15[2,3],ymm13[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm13, %zmm14, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm2, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm23 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm23, %zmm2, %zmm5
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm5 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm31, %ymm13 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm13 = ymm31[1],mem[1],ymm31[3],mem[3]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm15 = ymm19[1],ymm20[1],ymm19[3],ymm20[3]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm15[2,3],ymm13[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm13, %zmm5, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm20, %zmm2, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm2, %zmm14
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm14 {%k1}
+; AVX512BW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm13 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm15 = ymm26[1],ymm18[1],ymm26[3],ymm18[3]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm15[2,3],ymm13[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm13, %zmm14, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm2, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm14 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm14 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm13 = ymm29[1],ymm28[1],ymm29[3],ymm28[3]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm15 = ymm27[1],ymm30[1],ymm27[3],ymm30[3]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm13 = ymm15[2,3],ymm13[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm13, %zmm14, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm2, %zmm13
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm29 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm2, %zmm14
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm14 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm11 = ymm11[1],mem[1],ymm11[3],mem[3]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm9 = ymm9[1],ymm16[1],ymm9[3],ymm16[3]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm9 = ymm9[2,3],ymm11[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm9, %zmm14, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm2, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm10
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm10 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm10 {%k1}
+; AVX512BW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm7 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm7 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm5 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm5 = ymm8[1],mem[1],ymm8[3],mem[3]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm5 = ymm5[2,3],ymm7[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm5, %zmm10, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm2, %zmm6
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm3, %zmm8, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm2 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm3 = ymm17[1],ymm22[1],ymm17[3],ymm22[3]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm12[1],ymm0[3],ymm12[3]
+; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm3[2,3]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm2, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [4,12,4,12,4,12,4,12]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm13
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm11
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm2 = zmm22[0],zmm9[0],zmm22[2],zmm9[2],zmm22[4],zmm9[4],zmm22[6],zmm9[6]
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 3136(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 3072(%rdi), %zmm15
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 3264(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 3200(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm6 = <u,u,4,12>
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm6, %zmm4
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm3, %zmm2, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm21[0],zmm17[0],zmm21[2],zmm17[2],zmm21[4],zmm17[4],zmm21[6],zmm17[6]
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm21
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm6, %zmm4
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm12[0],zmm23[0],zmm12[2],zmm23[2],zmm12[4],zmm23[4],zmm12[6],zmm23[6]
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm6, %zmm4
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm20, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm1[0],zmm19[0],zmm1[2],zmm19[2],zmm1[4],zmm19[4],zmm1[6],zmm19[6]
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 1600(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1536(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 1728(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1664(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm6, %zmm4
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm1[0],zmm10[0],zmm1[2],zmm10[2],zmm1[4],zmm10[4],zmm1[6],zmm10[6]
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm6, %zmm4
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm0, %zmm2
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 = zmm30[0],zmm29[0],zmm30[2],zmm29[2],zmm30[4],zmm29[4],zmm30[6],zmm29[6]
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 2624(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2560(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 2752(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2688(%rdi), %zmm4
+; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm6, %zmm4
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2112(%rdi), %zmm7
+; AVX512BW-NEXT:    vmovdqa64 2048(%rdi), %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2240(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 2176(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm6, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm4 = zmm16[0],zmm30[0],zmm16[2],zmm30[2],zmm16[4],zmm30[4],zmm16[6],zmm30[6]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 3648(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 3584(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm3, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 3776(%rdi), %zmm3
+; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 3712(%rdi), %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm3, %zmm1, %zmm6
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
+; AVX512BW-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm1 # 64-byte Folded Reload
+; AVX512BW-NEXT:    # zmm1 = zmm8[0],mem[0],zmm8[2],mem[2],zmm8[4],mem[4],zmm8[6],mem[6]
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [5,13,5,13,5,13,5,13]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm0, %zmm15
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,5,13>
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm15[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm0, %zmm13
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm4 = zmm22[1],zmm9[1],zmm22[3],zmm9[3],zmm22[5],zmm9[5],zmm22[7],zmm9[7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm4 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm20, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 %zmm21, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm23 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm18 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm18, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, %zmm19
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm4 = zmm17[1],zmm19[1],zmm17[3],zmm19[3],zmm17[5],zmm19[5],zmm17[7],zmm19[7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm14, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm3 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm12, %zmm4 # 64-byte Folded Reload
+; AVX512BW-NEXT:    # zmm4 = zmm12[1],mem[1],zmm12[3],mem[3],zmm12[5],mem[5],zmm12[7],mem[7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm3 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm4 # 64-byte Folded Reload
+; AVX512BW-NEXT:    # zmm4 = zmm4[1],mem[1],zmm4[3],mem[3],zmm4[5],mem[5],zmm4[7],mem[7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm29 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm3 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm27
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm4 = zmm4[1],zmm10[1],zmm4[3],zmm10[3],zmm4[5],zmm10[5],zmm4[7],zmm10[7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm3 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm4 # 64-byte Folded Reload
+; AVX512BW-NEXT:    # zmm4 = zmm4[1],mem[1],zmm4[3],mem[3],zmm4[5],mem[5],zmm4[7],mem[7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm4 = zmm5[1],zmm30[1],zmm5[3],zmm30[3],zmm5[5],zmm30[5],zmm5[7],zmm30[7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm4 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm4, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm2
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm31, %zmm6, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm30 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm30, %zmm3, %zmm1
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm1 # 64-byte Folded Reload
+; AVX512BW-NEXT:    # zmm1 = zmm4[1],mem[1],zmm4[3],mem[3],zmm4[5],mem[5],zmm4[7],mem[7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1 {%k1}
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [6,14,6,14,6,14,6,14]
+; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = <u,u,6,14>
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm3
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm3 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm22, %zmm3 {%k1} # 64-byte Folded Reload
+; AVX512BW-NEXT:    # zmm3 {%k1} = zmm22[0],mem[0],zmm22[2],mem[2],zmm22[4],mem[4],zmm22[6],mem[6]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm20, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm21, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm17, %zmm3
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm0, %zmm3
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 {%k1} = zmm23[0],zmm18[0],zmm23[2],zmm18[2],zmm23[4],zmm18[4],zmm23[6],zmm18[6]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm16, %zmm0, %zmm14
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm14[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm18
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 {%k1} = zmm20[0],zmm14[0],zmm20[2],zmm14[2],zmm20[4],zmm14[4],zmm20[6],zmm14[6]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 {%k1} = zmm26[0],zmm25[0],zmm26[2],zmm25[2],zmm26[4],zmm25[4],zmm26[6],zmm25[6]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm29 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 {%k1} = zmm24[0],zmm29[0],zmm24[2],zmm29[2],zmm24[4],zmm29[4],zmm24[6],zmm29[6]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm1, %zmm3
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm19 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm3 {%k1} = zmm22[0],zmm19[0],zmm22[2],zmm19[2],zmm22[4],zmm19[4],zmm22[6],zmm19[6]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm1, %zmm8
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm0, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm5 {%k1} = zmm21[0],zmm17[0],zmm21[2],zmm17[2],zmm21[4],zmm17[4],zmm21[6],zmm17[6]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm5, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm4 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vpermi2q %zmm31, %zmm6, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm30, %zmm3, %zmm1
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm23 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} zmm4 {%k1} = zmm1[0],zmm23[0],zmm1[2],zmm23[2],zmm1[4],zmm23[4],zmm1[6],zmm23[6]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm4, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm30
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm5
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm16 = [7,15,7,15,7,15,7,15]
+; AVX512BW-NEXT:    # zmm16 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm16, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm0 {%k1} = zmm4[1],zmm1[1],zmm4[3],zmm1[3],zmm4[5],zmm1[5],zmm4[7],zmm1[7]
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm0
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm6
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm3 = [0,8,0,8,0,8,0,8]
+; AVX512BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm13 = [1,9,1,9,1,9,1,9]
+; AVX512BW-NEXT:    # zmm13 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm13, %zmm6
+; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm3, %zmm30
+; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm13, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm16, %zmm1
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm1 {%k1} = zmm20[1],zmm14[1],zmm20[3],zmm14[3],zmm20[5],zmm14[5],zmm20[7],zmm14[7]
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm3, %zmm20
+; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm13, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm3, %zmm18
+; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm28
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm16, %zmm12
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm12 {%k1} = zmm26[1],zmm25[1],zmm26[3],zmm25[3],zmm26[5],zmm25[5],zmm26[7],zmm25[7]
+; AVX512BW-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm3, %zmm26
+; AVX512BW-NEXT:    vmovdqu64 %zmm26, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm13, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm3, %zmm28
+; AVX512BW-NEXT:    vpermt2q %zmm15, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm25
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm16, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm7
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm1 {%k1} = zmm24[1],zmm29[1],zmm24[3],zmm29[3],zmm24[5],zmm29[5],zmm24[7],zmm29[7]
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm3, %zmm7
+; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm13, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm3, %zmm25
+; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm27
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm16, %zmm9
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm9 {%k1} = zmm22[1],zmm19[1],zmm22[3],zmm19[3],zmm22[5],zmm19[5],zmm22[7],zmm19[7]
+; AVX512BW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm3, %zmm22
+; AVX512BW-NEXT:    vmovdqu64 %zmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm13, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm3, %zmm27
+; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm31 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm29
+; AVX512BW-NEXT:    vmovdqa64 %zmm31, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm16, %zmm31
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm31 {%k1} = zmm21[1],zmm17[1],zmm21[3],zmm17[3],zmm21[5],zmm17[5],zmm21[7],zmm17[7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm19
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm3, %zmm19
+; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm13, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm3, %zmm29
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm22
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm3, %zmm22
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm23, %zmm3, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512BW-NEXT:    vpermi2q %zmm26, %zmm15, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm13, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm17
+; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm0
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm13, %zmm17
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm1
+; AVX512BW-NEXT:    vpermt2q %zmm23, %zmm13, %zmm1
+; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpermi2q %zmm26, %zmm15, %zmm13
+; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm16, %zmm15
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm15 {%k1} = zmm24[1],zmm23[1],zmm24[3],zmm23[3],zmm24[5],zmm23[5],zmm24[7],zmm23[7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm21
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm15
+; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm16, %zmm15
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} zmm15 {%k1} = zmm10[1],zmm8[1],zmm10[3],zmm8[3],zmm10[5],zmm8[5],zmm10[7],zmm8[7]
+; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm23
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm12 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm5 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm6 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm7 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm8 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm10 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm0 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm14 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} ymm16 = <u,u,7,15>
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm15 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm9 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm0 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm1 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm2 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm4 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm11 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512BW-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm11 # 64-byte Folded Reload
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm15 = ymm12[0,1,2,3],ymm15[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm15, %zmm12, %zmm12
+; AVX512BW-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm15 = ymm5[0,1,2,3],ymm9[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm15, %zmm5, %zmm5
+; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm15 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm15, %zmm0, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm15 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm15, %zmm0, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm15 = ymm8[0,1,2,3],ymm2[4,5,6,7]
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm15, %zmm0, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm15 = ymm10[0,1,2,3],ymm4[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm15, %zmm31, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vpblendd $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX512BW-NEXT:    # ymm15 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm15, %zmm21, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm15 = ymm14[0,1,2,3],ymm11[4,5,6,7]
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm15, %zmm23, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm18 {%k1}
+; AVX512BW-NEXT:    vmovdqa 192(%rdi), %xmm0
+; AVX512BW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512BW-NEXT:    vmovdqa 128(%rdi), %xmm1
+; AVX512BW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX512BW-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512BW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512BW-NEXT:    vmovdqa 64(%rdi), %xmm1
+; AVX512BW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm16 = xmm0[0],xmm1[0]
+; AVX512BW-NEXT:    vinserti32x4 $1, %xmm2, %ymm16, %ymm2
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm18, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm30 {%k1}
+; AVX512BW-NEXT:    vmovdqa 704(%rdi), %xmm8
+; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %xmm20
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm20[0],xmm8[0]
+; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %xmm18
+; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %xmm16
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm31 = xmm16[0],xmm18[0]
+; AVX512BW-NEXT:    vinserti32x4 $1, %xmm2, %ymm31, %ymm2
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm30, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm25 {%k1}
+; AVX512BW-NEXT:    vmovdqa 1216(%rdi), %xmm0
+; AVX512BW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512BW-NEXT:    vmovdqa 1152(%rdi), %xmm1
+; AVX512BW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm31 = xmm1[0],xmm0[0]
+; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %xmm23
+; AVX512BW-NEXT:    vmovdqa 1024(%rdi), %xmm14
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm30 = xmm14[0],xmm23[0]
+; AVX512BW-NEXT:    vinserti32x4 $1, %xmm31, %ymm30, %ymm30
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm30, %zmm25, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm28 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 1728(%rdi), %xmm30
+; AVX512BW-NEXT:    vmovdqa64 1664(%rdi), %xmm31
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm31[0],xmm30[0]
+; AVX512BW-NEXT:    vmovdqa 1600(%rdi), %xmm12
+; AVX512BW-NEXT:    vmovdqa 1536(%rdi), %xmm11
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm25 = xmm11[0],xmm12[0]
+; AVX512BW-NEXT:    vinserti32x4 $1, %xmm2, %ymm25, %ymm2
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm28, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm29 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 2240(%rdi), %xmm19
+; AVX512BW-NEXT:    vmovdqa64 2176(%rdi), %xmm21
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm25 = xmm21[0],xmm19[0]
+; AVX512BW-NEXT:    vmovdqa 2112(%rdi), %xmm15
+; AVX512BW-NEXT:    vmovdqa 2048(%rdi), %xmm1
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm28 = xmm1[0],xmm15[0]
+; AVX512BW-NEXT:    vinserti32x4 $1, %xmm25, %ymm28, %ymm25
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm25, %zmm29, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm27 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 2752(%rdi), %xmm28
+; AVX512BW-NEXT:    vmovdqa64 2688(%rdi), %xmm29
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm29[0],xmm28[0]
+; AVX512BW-NEXT:    vmovdqa 2624(%rdi), %xmm10
+; AVX512BW-NEXT:    vmovdqa 2560(%rdi), %xmm9
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm25 = xmm9[0],xmm10[0]
+; AVX512BW-NEXT:    vinserti32x4 $1, %xmm2, %ymm25, %ymm2
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm27, %zmm0
+; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm22 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 3264(%rdi), %xmm25
+; AVX512BW-NEXT:    vmovdqa64 3200(%rdi), %xmm27
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm4 = xmm27[0],xmm25[0]
+; AVX512BW-NEXT:    vmovdqa 3136(%rdi), %xmm0
+; AVX512BW-NEXT:    vmovdqa 3072(%rdi), %xmm2
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm26 = xmm2[0],xmm0[0]
+; AVX512BW-NEXT:    vinserti32x4 $1, %xmm4, %ymm26, %ymm4
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm4, %zmm22, %zmm22
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqa 3776(%rdi), %xmm7
+; AVX512BW-NEXT:    vmovdqa64 3712(%rdi), %xmm26
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm4 = xmm26[0],xmm7[0]
+; AVX512BW-NEXT:    vmovdqa 3648(%rdi), %xmm6
+; AVX512BW-NEXT:    vmovdqa 3584(%rdi), %xmm5
+; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm24 = xmm5[0],xmm6[0]
+; AVX512BW-NEXT:    vinserti32x4 $1, %xmm4, %ymm24, %ymm4
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm4, %zmm3, %zmm24
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm17 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm4 = xmm27[1],xmm25[1]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm4, %ymm2, %ymm2
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm2, %zmm17, %zmm2
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm0 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm3 = xmm20[1],xmm8[1]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm4 = xmm16[1],xmm18[1]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm3, %ymm4, %ymm3
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm3, %zmm0, %zmm3
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm0 {%k1}
+; AVX512BW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX512BW-NEXT:    # xmm4 = xmm4[1],mem[1]
+; AVX512BW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX512BW-NEXT:    # xmm8 = xmm8[1],mem[1]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm4, %ymm8, %ymm4
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm4, %zmm0, %zmm4
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm0 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm8 = xmm31[1],xmm30[1]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm11 = xmm11[1],xmm12[1]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm8, %ymm11, %ymm8
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm8, %zmm0, %zmm8
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm0 {%k1}
+; AVX512BW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX512BW-NEXT:    vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
+; AVX512BW-NEXT:    # xmm11 = xmm11[1],mem[1]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm12 = xmm14[1],xmm23[1]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm11, %ymm12, %ymm11
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm11, %zmm0, %zmm11
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm0 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm12 = xmm29[1],xmm28[1]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm9 = xmm9[1],xmm10[1]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm12, %ymm9, %ymm9
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm9, %zmm0, %zmm9
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm0 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm10 = xmm21[1],xmm19[1]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm12 = xmm1[1],xmm15[1]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm10, %ymm12, %ymm10
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm10, %zmm0, %zmm10
+; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm13 {%k1}
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm7 = xmm26[1],xmm7[1]
+; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm5 = xmm5[1],xmm6[1]
+; AVX512BW-NEXT:    vinserti128 $1, %xmm7, %ymm5, %ymm5
+; AVX512BW-NEXT:    vinserti64x4 $0, %ymm5, %zmm13, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 %zmm24, 448(%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm22, 384(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 320(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 256(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%rsi)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%rsi)
+; AVX512BW-NEXT:    vmovdqa64 %zmm1, 448(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm10, 256(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm9, 320(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm11, 128(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm8, 192(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm4, (%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm3, 64(%rdx)
+; AVX512BW-NEXT:    vmovdqa64 %zmm2, 384(%rdx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 448(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 256(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 320(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 384(%rcx)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 448(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 256(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 320(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 384(%r8)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 448(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 256(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 320(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%r9)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 384(%r9)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 448(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 256(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 320(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 384(%rax)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 448(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 256(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 320(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 384(%rax)
+; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX512BW-NEXT:    vmovups (%rsp), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 384(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 448(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 256(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 320(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 128(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 192(%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, (%rax)
+; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512BW-NEXT:    vmovaps %zmm0, 64(%rax)
+; AVX512BW-NEXT:    addq $6728, %rsp # imm = 0x1A48
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %wide.vec = load <512 x i64>, ptr %in.vec, align 64
+  %strided.vec0 = shufflevector <512 x i64> %wide.vec, <512 x i64> poison, <64 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120, i32 128, i32 136, i32 144, i32 152, i32 160, i32 168, i32 176, i32 184, i32 192, i32 200, i32 208, i32 216, i32 224, i32 232, i32 240, i32 248, i32 256, i32 264, i32 272, i32 280, i32 288, i32 296, i32 304, i32 312, i32 320, i32 328, i32 336, i32 344, i32 352, i32 360, i32 368, i32 376, i32 384, i32 392, i32 400, i32 408, i32 416, i32 424, i32 432, i32 440, i32 448, i32 456, i32 464, i32 472, i32 480, i32 488, i32 496, i32 504>
+  %strided.vec1 = shufflevector <512 x i64> %wide.vec, <512 x i64> poison, <64 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121, i32 129, i32 137, i32 145, i32 153, i32 161, i32 169, i32 177, i32 185, i32 193, i32 201, i32 209, i32 217, i32 225, i32 233, i32 241, i32 249, i32 257, i32 265, i32 273, i32 281, i32 289, i32 297, i32 305, i32 313, i32 321, i32 329, i32 337, i32 345, i32 353, i32 361, i32 369, i32 377, i32 385, i32 393, i32 401, i32 409, i32 417, i32 425, i32 433, i32 441, i32 449, i32 457, i32 465, i32 473, i32 481, i32 489, i32 497, i32 505>
+  %strided.vec2 = shufflevector <512 x i64> %wide.vec, <512 x i64> poison, <64 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122, i32 130, i32 138, i32 146, i32 154, i32 162, i32 170, i32 178, i32 186, i32 194, i32 202, i32 210, i32 218, i32 226, i32 234, i32 242, i32 250, i32 258, i32 266, i32 274, i32 282, i32 290, i32 298, i32 306, i32 314, i32 322, i32 330, i32 338, i32 346, i32 354, i32 362, i32 370, i32 378, i32 386, i32 394, i32 402, i32 410, i32 418, i32 426, i32 434, i32 442, i32 450, i32 458, i32 466, i32 474, i32 482, i32 490, i32 498, i32 506>
+  %strided.vec3 = shufflevector <512 x i64> %wide.vec, <512 x i64> poison, <64 x i32> <i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 51, i32 59, i32 67, i32 75, i32 83, i32 91, i32 99, i32 107, i32 115, i32 123, i32 131, i32 139, i32 147, i32 155, i32 163, i32 171, i32 179, i32 187, i32 195, i32 203, i32 211, i32 219, i32 227, i32 235, i32 243, i32 251, i32 259, i32 267, i32 275, i32 283, i32 291, i32 299, i32 307, i32 315, i32 323, i32 331, i32 339, i32 347, i32 355, i32 363, i32 371, i32 379, i32 387, i32 395, i32 403, i32 411, i32 419, i32 427, i32 435, i32 443, i32 451, i32 459, i32 467, i32 475, i32 483, i32 491, i32 499, i32 507>
+  %strided.vec4 = shufflevector <512 x i64> %wide.vec, <512 x i64> poison, <64 x i32> <i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 52, i32 60, i32 68, i32 76, i32 84, i32 92, i32 100, i32 108, i32 116, i32 124, i32 132, i32 140, i32 148, i32 156, i32 164, i32 172, i32 180, i32 188, i32 196, i32 204, i32 212, i32 220, i32 228, i32 236, i32 244, i32 252, i32 260, i32 268, i32 276, i32 284, i32 292, i32 300, i32 308, i32 316, i32 324, i32 332, i32 340, i32 348, i32 356, i32 364, i32 372, i32 380, i32 388, i32 396, i32 404, i32 412, i32 420, i32 428, i32 436, i32 444, i32 452, i32 460, i32 468, i32 476, i32 484, i32 492, i32 500, i32 508>
+  %strided.vec5 = shufflevector <512 x i64> %wide.vec, <512 x i64> poison, <64 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125, i32 133, i32 141, i32 149, i32 157, i32 165, i32 173, i32 181, i32 189, i32 197, i32 205, i32 213, i32 221, i32 229, i32 237, i32 245, i32 253, i32 261, i32 269, i32 277, i32 285, i32 293, i32 301, i32 309, i32 317, i32 325, i32 333, i32 341, i32 349, i32 357, i32 365, i32 373, i32 381, i32 389, i32 397, i32 405, i32 413, i32 421, i32 429, i32 437, i32 445, i32 453, i32 461, i32 469, i32 477, i32 485, i32 493, i32 501, i32 509>
+  %strided.vec6 = shufflevector <512 x i64> %wide.vec, <512 x i64> poison, <64 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126, i32 134, i32 142, i32 150, i32 158, i32 166, i32 174, i32 182, i32 190, i32 198, i32 206, i32 214, i32 222, i32 230, i32 238, i32 246, i32 254, i32 262, i32 270, i32 278, i32 286, i32 294, i32 302, i32 310, i32 318, i32 326, i32 334, i32 342, i32 350, i32 358, i32 366, i32 374, i32 382, i32 390, i32 398, i32 406, i32 414, i32 422, i32 430, i32 438, i32 446, i32 454, i32 462, i32 470, i32 478, i32 486, i32 494, i32 502, i32 510>
+  %strided.vec7 = shufflevector <512 x i64> %wide.vec, <512 x i64> poison, <64 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127, i32 135, i32 143, i32 151, i32 159, i32 167, i32 175, i32 183, i32 191, i32 199, i32 207, i32 215, i32 223, i32 231, i32 239, i32 247, i32 255, i32 263, i32 271, i32 279, i32 287, i32 295, i32 303, i32 311, i32 319, i32 327, i32 335, i32 343, i32 351, i32 359, i32 367, i32 375, i32 383, i32 391, i32 399, i32 407, i32 415, i32 423, i32 431, i32 439, i32 447, i32 455, i32 463, i32 471, i32 479, i32 487, i32 495, i32 503, i32 511>
+  store <64 x i64> %strided.vec0, ptr %out.vec0, align 64
+  store <64 x i64> %strided.vec1, ptr %out.vec1, align 64
+  store <64 x i64> %strided.vec2, ptr %out.vec2, align 64
+  store <64 x i64> %strided.vec3, ptr %out.vec3, align 64
+  store <64 x i64> %strided.vec4, ptr %out.vec4, align 64
+  store <64 x i64> %strided.vec5, ptr %out.vec5, align 64
+  store <64 x i64> %strided.vec6, ptr %out.vec6, align 64
+  store <64 x i64> %strided.vec7, ptr %out.vec7, align 64
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-2.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-2.ll
index 7f8be0e3..686c9a6 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-2.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-2.ll
@@ -57,13 +57,10 @@ define void @load_i8_stride2_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) noun
 ; AVX512BW-NEXT:    vpextrw $0, %xmm0, (%rdx)
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <4 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <4 x i8> %wide.vec, <4 x i8> poison, <2 x i32> <i32 0, i32 2>
   %strided.vec1 = shufflevector <4 x i8> %wide.vec, <4 x i8> poison, <2 x i32> <i32 1, i32 3>
-
   store <2 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i8> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -107,13 +104,10 @@ define void @load_i8_stride2_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) noun
 ; AVX512BW-NEXT:    vmovd %xmm0, (%rdx)
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <8 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <8 x i8> %wide.vec, <8 x i8> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
   %strided.vec1 = shufflevector <8 x i8> %wide.vec, <8 x i8> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
-
   store <4 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i8> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -156,13 +150,10 @@ define void @load_i8_stride2_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) noun
 ; AVX512BW-NEXT:    vmovq %xmm1, (%rdx)
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <16 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <16 x i8> %wide.vec, <16 x i8> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %strided.vec1 = shufflevector <16 x i8> %wide.vec, <16 x i8> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
-
   store <8 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i8> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -229,13 +220,10 @@ define void @load_i8_stride2_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nou
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <32 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <32 x i8> %wide.vec, <32 x i8> poison, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %strided.vec1 = shufflevector <32 x i8> %wide.vec, <32 x i8> poison, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
-
   store <16 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i8> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -373,13 +361,10 @@ define void @load_i8_stride2_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nou
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <64 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46, i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62>
   %strided.vec1 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <32 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31, i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47, i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63>
-
   store <32 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i8> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 
@@ -593,13 +578,10 @@ define void @load_i8_stride2_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nou
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <128 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <128 x i8> %wide.vec, <128 x i8> poison, <64 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46, i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62, i32 64, i32 66, i32 68, i32 70, i32 72, i32 74, i32 76, i32 78, i32 80, i32 82, i32 84, i32 86, i32 88, i32 90, i32 92, i32 94, i32 96, i32 98, i32 100, i32 102, i32 104, i32 106, i32 108, i32 110, i32 112, i32 114, i32 116, i32 118, i32 120, i32 122, i32 124, i32 126>
   %strided.vec1 = shufflevector <128 x i8> %wide.vec, <128 x i8> poison, <64 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31, i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47, i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63, i32 65, i32 67, i32 69, i32 71, i32 73, i32 75, i32 77, i32 79, i32 81, i32 83, i32 85, i32 87, i32 89, i32 91, i32 93, i32 95, i32 97, i32 99, i32 101, i32 103, i32 105, i32 107, i32 109, i32 111, i32 113, i32 115, i32 117, i32 119, i32 121, i32 123, i32 125, i32 127>
-
   store <64 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i8> %strided.vec1, ptr %out.vec1, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-3.ll
index baffb9c..21b6e38 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-3.ll
@@ -48,15 +48,12 @@ define void @load_i8_stride3_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX-NEXT:    vpextrw $0, %xmm0, (%rcx)
 ; AVX-NEXT:    retq
   %wide.vec = load <6 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <6 x i8> %wide.vec, <6 x i8> poison, <2 x i32> <i32 0, i32 3>
   %strided.vec1 = shufflevector <6 x i8> %wide.vec, <6 x i8> poison, <2 x i32> <i32 1, i32 4>
   %strided.vec2 = shufflevector <6 x i8> %wide.vec, <6 x i8> poison, <2 x i32> <i32 2, i32 5>
-
   store <2 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i8> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -106,15 +103,12 @@ define void @load_i8_stride3_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX-NEXT:    vmovd %xmm0, (%rcx)
 ; AVX-NEXT:    retq
   %wide.vec = load <12 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <12 x i8> %wide.vec, <12 x i8> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
   %strided.vec1 = shufflevector <12 x i8> %wide.vec, <12 x i8> poison, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
   %strided.vec2 = shufflevector <12 x i8> %wide.vec, <12 x i8> poison, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
-
   store <4 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i8> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -196,15 +190,12 @@ define void @load_i8_stride3_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX-NEXT:    vmovq %xmm0, (%rcx)
 ; AVX-NEXT:    retq
   %wide.vec = load <24 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <24 x i8> %wide.vec, <24 x i8> poison, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
   %strided.vec1 = shufflevector <24 x i8> %wide.vec, <24 x i8> poison, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
   %strided.vec2 = shufflevector <24 x i8> %wide.vec, <24 x i8> poison, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
-
   store <8 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i8> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -392,15 +383,12 @@ define void @load_i8_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vmovdqa %xmm1, (%rcx)
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <48 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <48 x i8> %wide.vec, <48 x i8> poison, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
   %strided.vec1 = shufflevector <48 x i8> %wide.vec, <48 x i8> poison, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
   %strided.vec2 = shufflevector <48 x i8> %wide.vec, <48 x i8> poison, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
-
   store <16 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i8> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -764,15 +752,12 @@ define void @load_i8_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <96 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <96 x i8> %wide.vec, <96 x i8> poison, <32 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45, i32 48, i32 51, i32 54, i32 57, i32 60, i32 63, i32 66, i32 69, i32 72, i32 75, i32 78, i32 81, i32 84, i32 87, i32 90, i32 93>
   %strided.vec1 = shufflevector <96 x i8> %wide.vec, <96 x i8> poison, <32 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46, i32 49, i32 52, i32 55, i32 58, i32 61, i32 64, i32 67, i32 70, i32 73, i32 76, i32 79, i32 82, i32 85, i32 88, i32 91, i32 94>
   %strided.vec2 = shufflevector <96 x i8> %wide.vec, <96 x i8> poison, <32 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47, i32 50, i32 53, i32 56, i32 59, i32 62, i32 65, i32 68, i32 71, i32 74, i32 77, i32 80, i32 83, i32 86, i32 89, i32 92, i32 95>
-
   store <32 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i8> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 
@@ -1507,15 +1492,12 @@ define void @load_i8_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <192 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <192 x i8> %wide.vec, <192 x i8> poison, <64 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45, i32 48, i32 51, i32 54, i32 57, i32 60, i32 63, i32 66, i32 69, i32 72, i32 75, i32 78, i32 81, i32 84, i32 87, i32 90, i32 93, i32 96, i32 99, i32 102, i32 105, i32 108, i32 111, i32 114, i32 117, i32 120, i32 123, i32 126, i32 129, i32 132, i32 135, i32 138, i32 141, i32 144, i32 147, i32 150, i32 153, i32 156, i32 159, i32 162, i32 165, i32 168, i32 171, i32 174, i32 177, i32 180, i32 183, i32 186, i32 189>
   %strided.vec1 = shufflevector <192 x i8> %wide.vec, <192 x i8> poison, <64 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46, i32 49, i32 52, i32 55, i32 58, i32 61, i32 64, i32 67, i32 70, i32 73, i32 76, i32 79, i32 82, i32 85, i32 88, i32 91, i32 94, i32 97, i32 100, i32 103, i32 106, i32 109, i32 112, i32 115, i32 118, i32 121, i32 124, i32 127, i32 130, i32 133, i32 136, i32 139, i32 142, i32 145, i32 148, i32 151, i32 154, i32 157, i32 160, i32 163, i32 166, i32 169, i32 172, i32 175, i32 178, i32 181, i32 184, i32 187, i32 190>
   %strided.vec2 = shufflevector <192 x i8> %wide.vec, <192 x i8> poison, <64 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47, i32 50, i32 53, i32 56, i32 59, i32 62, i32 65, i32 68, i32 71, i32 74, i32 77, i32 80, i32 83, i32 86, i32 89, i32 92, i32 95, i32 98, i32 101, i32 104, i32 107, i32 110, i32 113, i32 116, i32 119, i32 122, i32 125, i32 128, i32 131, i32 134, i32 137, i32 140, i32 143, i32 146, i32 149, i32 152, i32 155, i32 158, i32 161, i32 164, i32 167, i32 170, i32 173, i32 176, i32 179, i32 182, i32 185, i32 188, i32 191>
-
   store <64 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i8> %strided.vec2, ptr %out.vec2, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
index 4c7ffc5..5a297c8 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
@@ -70,17 +70,14 @@ define void @load_i8_stride4_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vpextrw $0, %xmm0, (%r8)
 ; AVX512-NEXT:    retq
   %wide.vec = load <8 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <8 x i8> %wide.vec, <8 x i8> poison, <2 x i32> <i32 0, i32 4>
   %strided.vec1 = shufflevector <8 x i8> %wide.vec, <8 x i8> poison, <2 x i32> <i32 1, i32 5>
   %strided.vec2 = shufflevector <8 x i8> %wide.vec, <8 x i8> poison, <2 x i32> <i32 2, i32 6>
   %strided.vec3 = shufflevector <8 x i8> %wide.vec, <8 x i8> poison, <2 x i32> <i32 3, i32 7>
-
   store <2 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <2 x i8> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -145,17 +142,14 @@ define void @load_i8_stride4_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vmovd %xmm3, (%r8)
 ; AVX512-NEXT:    retq
   %wide.vec = load <16 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <16 x i8> %wide.vec, <16 x i8> poison, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
   %strided.vec1 = shufflevector <16 x i8> %wide.vec, <16 x i8> poison, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
   %strided.vec2 = shufflevector <16 x i8> %wide.vec, <16 x i8> poison, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
   %strided.vec3 = shufflevector <16 x i8> %wide.vec, <16 x i8> poison, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
-
   store <4 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <4 x i8> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -272,17 +266,14 @@ define void @load_i8_stride4_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <32 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <32 x i8> %wide.vec, <32 x i8> poison, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
   %strided.vec1 = shufflevector <32 x i8> %wide.vec, <32 x i8> poison, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
   %strided.vec2 = shufflevector <32 x i8> %wide.vec, <32 x i8> poison, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
   %strided.vec3 = shufflevector <32 x i8> %wide.vec, <32 x i8> poison, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
-
   store <8 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <8 x i8> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -536,17 +527,14 @@ define void @load_i8_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <64 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
   %strided.vec1 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
   %strided.vec2 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
   %strided.vec3 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
-
   store <16 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <16 x i8> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -1060,17 +1048,14 @@ define void @load_i8_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <128 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <128 x i8> %wide.vec, <128 x i8> poison, <32 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124>
   %strided.vec1 = shufflevector <128 x i8> %wide.vec, <128 x i8> poison, <32 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125>
   %strided.vec2 = shufflevector <128 x i8> %wide.vec, <128 x i8> poison, <32 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126>
   %strided.vec3 = shufflevector <128 x i8> %wide.vec, <128 x i8> poison, <32 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127>
-
   store <32 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <32 x i8> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 
@@ -2165,17 +2150,14 @@ define void @load_i8_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <256 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <256 x i8> %wide.vec, <256 x i8> poison, <64 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124, i32 128, i32 132, i32 136, i32 140, i32 144, i32 148, i32 152, i32 156, i32 160, i32 164, i32 168, i32 172, i32 176, i32 180, i32 184, i32 188, i32 192, i32 196, i32 200, i32 204, i32 208, i32 212, i32 216, i32 220, i32 224, i32 228, i32 232, i32 236, i32 240, i32 244, i32 248, i32 252>
   %strided.vec1 = shufflevector <256 x i8> %wide.vec, <256 x i8> poison, <64 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125, i32 129, i32 133, i32 137, i32 141, i32 145, i32 149, i32 153, i32 157, i32 161, i32 165, i32 169, i32 173, i32 177, i32 181, i32 185, i32 189, i32 193, i32 197, i32 201, i32 205, i32 209, i32 213, i32 217, i32 221, i32 225, i32 229, i32 233, i32 237, i32 241, i32 245, i32 249, i32 253>
   %strided.vec2 = shufflevector <256 x i8> %wide.vec, <256 x i8> poison, <64 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126, i32 130, i32 134, i32 138, i32 142, i32 146, i32 150, i32 154, i32 158, i32 162, i32 166, i32 170, i32 174, i32 178, i32 182, i32 186, i32 190, i32 194, i32 198, i32 202, i32 206, i32 210, i32 214, i32 218, i32 222, i32 226, i32 230, i32 234, i32 238, i32 242, i32 246, i32 250, i32 254>
   %strided.vec3 = shufflevector <256 x i8> %wide.vec, <256 x i8> poison, <64 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127, i32 131, i32 135, i32 139, i32 143, i32 147, i32 151, i32 155, i32 159, i32 163, i32 167, i32 171, i32 175, i32 179, i32 183, i32 187, i32 191, i32 195, i32 199, i32 203, i32 207, i32 211, i32 215, i32 219, i32 223, i32 227, i32 231, i32 235, i32 239, i32 243, i32 247, i32 251, i32 255>
-
   store <64 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <64 x i8> %strided.vec3, ptr %out.vec3, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
index 227caea..5103dc0 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
@@ -66,19 +66,16 @@ define void @load_i8_stride5_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX-NEXT:    vpextrw $0, %xmm0, (%r9)
 ; AVX-NEXT:    retq
   %wide.vec = load <10 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <10 x i8> %wide.vec, <10 x i8> poison, <2 x i32> <i32 0, i32 5>
   %strided.vec1 = shufflevector <10 x i8> %wide.vec, <10 x i8> poison, <2 x i32> <i32 1, i32 6>
   %strided.vec2 = shufflevector <10 x i8> %wide.vec, <10 x i8> poison, <2 x i32> <i32 2, i32 7>
   %strided.vec3 = shufflevector <10 x i8> %wide.vec, <10 x i8> poison, <2 x i32> <i32 3, i32 8>
   %strided.vec4 = shufflevector <10 x i8> %wide.vec, <10 x i8> poison, <2 x i32> <i32 4, i32 9>
-
   store <2 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <2 x i8> %strided.vec3, ptr %out.vec3, align 64
   store <2 x i8> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -174,19 +171,16 @@ define void @load_i8_stride5_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX-NEXT:    vmovd %xmm0, (%r9)
 ; AVX-NEXT:    retq
   %wide.vec = load <20 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <20 x i8> %wide.vec, <20 x i8> poison, <4 x i32> <i32 0, i32 5, i32 10, i32 15>
   %strided.vec1 = shufflevector <20 x i8> %wide.vec, <20 x i8> poison, <4 x i32> <i32 1, i32 6, i32 11, i32 16>
   %strided.vec2 = shufflevector <20 x i8> %wide.vec, <20 x i8> poison, <4 x i32> <i32 2, i32 7, i32 12, i32 17>
   %strided.vec3 = shufflevector <20 x i8> %wide.vec, <20 x i8> poison, <4 x i32> <i32 3, i32 8, i32 13, i32 18>
   %strided.vec4 = shufflevector <20 x i8> %wide.vec, <20 x i8> poison, <4 x i32> <i32 4, i32 9, i32 14, i32 19>
-
   store <4 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <4 x i8> %strided.vec3, ptr %out.vec3, align 64
   store <4 x i8> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -391,19 +385,16 @@ define void @load_i8_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX2-NEXT:    vmovq %xmm0, (%r9)
 ; AVX2-NEXT:    retq
   %wide.vec = load <40 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <40 x i8> %wide.vec, <40 x i8> poison, <8 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35>
   %strided.vec1 = shufflevector <40 x i8> %wide.vec, <40 x i8> poison, <8 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36>
   %strided.vec2 = shufflevector <40 x i8> %wide.vec, <40 x i8> poison, <8 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37>
   %strided.vec3 = shufflevector <40 x i8> %wide.vec, <40 x i8> poison, <8 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38>
   %strided.vec4 = shufflevector <40 x i8> %wide.vec, <40 x i8> poison, <8 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39>
-
   store <8 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <8 x i8> %strided.vec3, ptr %out.vec3, align 64
   store <8 x i8> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -960,19 +951,16 @@ define void @load_i8_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vmovdqa %xmm0, (%r9)
 ; AVX512-NEXT:    retq
   %wide.vec = load <80 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <80 x i8> %wide.vec, <80 x i8> poison, <16 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75>
   %strided.vec1 = shufflevector <80 x i8> %wide.vec, <80 x i8> poison, <16 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76>
   %strided.vec2 = shufflevector <80 x i8> %wide.vec, <80 x i8> poison, <16 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77>
   %strided.vec3 = shufflevector <80 x i8> %wide.vec, <80 x i8> poison, <16 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78>
   %strided.vec4 = shufflevector <80 x i8> %wide.vec, <80 x i8> poison, <16 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79>
-
   store <16 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <16 x i8> %strided.vec3, ptr %out.vec3, align 64
   store <16 x i8> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -2012,19 +2000,16 @@ define void @load_i8_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <160 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <160 x i8> %wide.vec, <160 x i8> poison, <32 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75, i32 80, i32 85, i32 90, i32 95, i32 100, i32 105, i32 110, i32 115, i32 120, i32 125, i32 130, i32 135, i32 140, i32 145, i32 150, i32 155>
   %strided.vec1 = shufflevector <160 x i8> %wide.vec, <160 x i8> poison, <32 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76, i32 81, i32 86, i32 91, i32 96, i32 101, i32 106, i32 111, i32 116, i32 121, i32 126, i32 131, i32 136, i32 141, i32 146, i32 151, i32 156>
   %strided.vec2 = shufflevector <160 x i8> %wide.vec, <160 x i8> poison, <32 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77, i32 82, i32 87, i32 92, i32 97, i32 102, i32 107, i32 112, i32 117, i32 122, i32 127, i32 132, i32 137, i32 142, i32 147, i32 152, i32 157>
   %strided.vec3 = shufflevector <160 x i8> %wide.vec, <160 x i8> poison, <32 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78, i32 83, i32 88, i32 93, i32 98, i32 103, i32 108, i32 113, i32 118, i32 123, i32 128, i32 133, i32 138, i32 143, i32 148, i32 153, i32 158>
   %strided.vec4 = shufflevector <160 x i8> %wide.vec, <160 x i8> poison, <32 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79, i32 84, i32 89, i32 94, i32 99, i32 104, i32 109, i32 114, i32 119, i32 124, i32 129, i32 134, i32 139, i32 144, i32 149, i32 154, i32 159>
-
   store <32 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <32 x i8> %strided.vec3, ptr %out.vec3, align 64
   store <32 x i8> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 
@@ -4239,19 +4224,16 @@ define void @load_i8_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <320 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <320 x i8> %wide.vec, <320 x i8> poison, <64 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75, i32 80, i32 85, i32 90, i32 95, i32 100, i32 105, i32 110, i32 115, i32 120, i32 125, i32 130, i32 135, i32 140, i32 145, i32 150, i32 155, i32 160, i32 165, i32 170, i32 175, i32 180, i32 185, i32 190, i32 195, i32 200, i32 205, i32 210, i32 215, i32 220, i32 225, i32 230, i32 235, i32 240, i32 245, i32 250, i32 255, i32 260, i32 265, i32 270, i32 275, i32 280, i32 285, i32 290, i32 295, i32 300, i32 305, i32 310, i32 315>
   %strided.vec1 = shufflevector <320 x i8> %wide.vec, <320 x i8> poison, <64 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76, i32 81, i32 86, i32 91, i32 96, i32 101, i32 106, i32 111, i32 116, i32 121, i32 126, i32 131, i32 136, i32 141, i32 146, i32 151, i32 156, i32 161, i32 166, i32 171, i32 176, i32 181, i32 186, i32 191, i32 196, i32 201, i32 206, i32 211, i32 216, i32 221, i32 226, i32 231, i32 236, i32 241, i32 246, i32 251, i32 256, i32 261, i32 266, i32 271, i32 276, i32 281, i32 286, i32 291, i32 296, i32 301, i32 306, i32 311, i32 316>
   %strided.vec2 = shufflevector <320 x i8> %wide.vec, <320 x i8> poison, <64 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77, i32 82, i32 87, i32 92, i32 97, i32 102, i32 107, i32 112, i32 117, i32 122, i32 127, i32 132, i32 137, i32 142, i32 147, i32 152, i32 157, i32 162, i32 167, i32 172, i32 177, i32 182, i32 187, i32 192, i32 197, i32 202, i32 207, i32 212, i32 217, i32 222, i32 227, i32 232, i32 237, i32 242, i32 247, i32 252, i32 257, i32 262, i32 267, i32 272, i32 277, i32 282, i32 287, i32 292, i32 297, i32 302, i32 307, i32 312, i32 317>
   %strided.vec3 = shufflevector <320 x i8> %wide.vec, <320 x i8> poison, <64 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78, i32 83, i32 88, i32 93, i32 98, i32 103, i32 108, i32 113, i32 118, i32 123, i32 128, i32 133, i32 138, i32 143, i32 148, i32 153, i32 158, i32 163, i32 168, i32 173, i32 178, i32 183, i32 188, i32 193, i32 198, i32 203, i32 208, i32 213, i32 218, i32 223, i32 228, i32 233, i32 238, i32 243, i32 248, i32 253, i32 258, i32 263, i32 268, i32 273, i32 278, i32 283, i32 288, i32 293, i32 298, i32 303, i32 308, i32 313, i32 318>
   %strided.vec4 = shufflevector <320 x i8> %wide.vec, <320 x i8> poison, <64 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79, i32 84, i32 89, i32 94, i32 99, i32 104, i32 109, i32 114, i32 119, i32 124, i32 129, i32 134, i32 139, i32 144, i32 149, i32 154, i32 159, i32 164, i32 169, i32 174, i32 179, i32 184, i32 189, i32 194, i32 199, i32 204, i32 209, i32 214, i32 219, i32 224, i32 229, i32 234, i32 239, i32 244, i32 249, i32 254, i32 259, i32 264, i32 269, i32 274, i32 279, i32 284, i32 289, i32 294, i32 299, i32 304, i32 309, i32 314, i32 319>
-
   store <64 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <64 x i8> %strided.vec3, ptr %out.vec3, align 64
   store <64 x i8> %strided.vec4, ptr %out.vec4, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
index c169b52..8cdb27a 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
@@ -78,21 +78,18 @@ define void @load_i8_stride6_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX-NEXT:    vpextrw $0, %xmm0, (%rax)
 ; AVX-NEXT:    retq
   %wide.vec = load <12 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <12 x i8> %wide.vec, <12 x i8> poison, <2 x i32> <i32 0, i32 6>
   %strided.vec1 = shufflevector <12 x i8> %wide.vec, <12 x i8> poison, <2 x i32> <i32 1, i32 7>
   %strided.vec2 = shufflevector <12 x i8> %wide.vec, <12 x i8> poison, <2 x i32> <i32 2, i32 8>
   %strided.vec3 = shufflevector <12 x i8> %wide.vec, <12 x i8> poison, <2 x i32> <i32 3, i32 9>
   %strided.vec4 = shufflevector <12 x i8> %wide.vec, <12 x i8> poison, <2 x i32> <i32 4, i32 10>
   %strided.vec5 = shufflevector <12 x i8> %wide.vec, <12 x i8> poison, <2 x i32> <i32 5, i32 11>
-
   store <2 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <2 x i8> %strided.vec3, ptr %out.vec3, align 64
   store <2 x i8> %strided.vec4, ptr %out.vec4, align 64
   store <2 x i8> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -201,21 +198,18 @@ define void @load_i8_stride6_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX-NEXT:    vmovd %xmm0, (%rax)
 ; AVX-NEXT:    retq
   %wide.vec = load <24 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <24 x i8> %wide.vec, <24 x i8> poison, <4 x i32> <i32 0, i32 6, i32 12, i32 18>
   %strided.vec1 = shufflevector <24 x i8> %wide.vec, <24 x i8> poison, <4 x i32> <i32 1, i32 7, i32 13, i32 19>
   %strided.vec2 = shufflevector <24 x i8> %wide.vec, <24 x i8> poison, <4 x i32> <i32 2, i32 8, i32 14, i32 20>
   %strided.vec3 = shufflevector <24 x i8> %wide.vec, <24 x i8> poison, <4 x i32> <i32 3, i32 9, i32 15, i32 21>
   %strided.vec4 = shufflevector <24 x i8> %wide.vec, <24 x i8> poison, <4 x i32> <i32 4, i32 10, i32 16, i32 22>
   %strided.vec5 = shufflevector <24 x i8> %wide.vec, <24 x i8> poison, <4 x i32> <i32 5, i32 11, i32 17, i32 23>
-
   store <4 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <4 x i8> %strided.vec3, ptr %out.vec3, align 64
   store <4 x i8> %strided.vec4, ptr %out.vec4, align 64
   store <4 x i8> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -457,21 +451,18 @@ define void @load_i8_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
   %wide.vec = load <48 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <48 x i8> %wide.vec, <48 x i8> poison, <8 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42>
   %strided.vec1 = shufflevector <48 x i8> %wide.vec, <48 x i8> poison, <8 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43>
   %strided.vec2 = shufflevector <48 x i8> %wide.vec, <48 x i8> poison, <8 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44>
   %strided.vec3 = shufflevector <48 x i8> %wide.vec, <48 x i8> poison, <8 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45>
   %strided.vec4 = shufflevector <48 x i8> %wide.vec, <48 x i8> poison, <8 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46>
   %strided.vec5 = shufflevector <48 x i8> %wide.vec, <48 x i8> poison, <8 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47>
-
   store <8 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <8 x i8> %strided.vec3, ptr %out.vec3, align 64
   store <8 x i8> %strided.vec4, ptr %out.vec4, align 64
   store <8 x i8> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -1675,21 +1666,18 @@ define void @load_i8_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vmovdqa %xmm0, (%rax)
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <96 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <96 x i8> %wide.vec, <96 x i8> poison, <16 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90>
   %strided.vec1 = shufflevector <96 x i8> %wide.vec, <96 x i8> poison, <16 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91>
   %strided.vec2 = shufflevector <96 x i8> %wide.vec, <96 x i8> poison, <16 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92>
   %strided.vec3 = shufflevector <96 x i8> %wide.vec, <96 x i8> poison, <16 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93>
   %strided.vec4 = shufflevector <96 x i8> %wide.vec, <96 x i8> poison, <16 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94>
   %strided.vec5 = shufflevector <96 x i8> %wide.vec, <96 x i8> poison, <16 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95>
-
   store <16 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <16 x i8> %strided.vec3, ptr %out.vec3, align 64
   store <16 x i8> %strided.vec4, ptr %out.vec4, align 64
   store <16 x i8> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -2905,21 +2893,18 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <192 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <192 x i8> %wide.vec, <192 x i8> poison, <32 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90, i32 96, i32 102, i32 108, i32 114, i32 120, i32 126, i32 132, i32 138, i32 144, i32 150, i32 156, i32 162, i32 168, i32 174, i32 180, i32 186>
   %strided.vec1 = shufflevector <192 x i8> %wide.vec, <192 x i8> poison, <32 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91, i32 97, i32 103, i32 109, i32 115, i32 121, i32 127, i32 133, i32 139, i32 145, i32 151, i32 157, i32 163, i32 169, i32 175, i32 181, i32 187>
   %strided.vec2 = shufflevector <192 x i8> %wide.vec, <192 x i8> poison, <32 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92, i32 98, i32 104, i32 110, i32 116, i32 122, i32 128, i32 134, i32 140, i32 146, i32 152, i32 158, i32 164, i32 170, i32 176, i32 182, i32 188>
   %strided.vec3 = shufflevector <192 x i8> %wide.vec, <192 x i8> poison, <32 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93, i32 99, i32 105, i32 111, i32 117, i32 123, i32 129, i32 135, i32 141, i32 147, i32 153, i32 159, i32 165, i32 171, i32 177, i32 183, i32 189>
   %strided.vec4 = shufflevector <192 x i8> %wide.vec, <192 x i8> poison, <32 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94, i32 100, i32 106, i32 112, i32 118, i32 124, i32 130, i32 136, i32 142, i32 148, i32 154, i32 160, i32 166, i32 172, i32 178, i32 184, i32 190>
   %strided.vec5 = shufflevector <192 x i8> %wide.vec, <192 x i8> poison, <32 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95, i32 101, i32 107, i32 113, i32 119, i32 125, i32 131, i32 137, i32 143, i32 149, i32 155, i32 161, i32 167, i32 173, i32 179, i32 185, i32 191>
-
   store <32 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <32 x i8> %strided.vec3, ptr %out.vec3, align 64
   store <32 x i8> %strided.vec4, ptr %out.vec4, align 64
   store <32 x i8> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 
@@ -5681,21 +5666,18 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <384 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <384 x i8> %wide.vec, <384 x i8> poison, <64 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90, i32 96, i32 102, i32 108, i32 114, i32 120, i32 126, i32 132, i32 138, i32 144, i32 150, i32 156, i32 162, i32 168, i32 174, i32 180, i32 186, i32 192, i32 198, i32 204, i32 210, i32 216, i32 222, i32 228, i32 234, i32 240, i32 246, i32 252, i32 258, i32 264, i32 270, i32 276, i32 282, i32 288, i32 294, i32 300, i32 306, i32 312, i32 318, i32 324, i32 330, i32 336, i32 342, i32 348, i32 354, i32 360, i32 366, i32 372, i32 378>
   %strided.vec1 = shufflevector <384 x i8> %wide.vec, <384 x i8> poison, <64 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91, i32 97, i32 103, i32 109, i32 115, i32 121, i32 127, i32 133, i32 139, i32 145, i32 151, i32 157, i32 163, i32 169, i32 175, i32 181, i32 187, i32 193, i32 199, i32 205, i32 211, i32 217, i32 223, i32 229, i32 235, i32 241, i32 247, i32 253, i32 259, i32 265, i32 271, i32 277, i32 283, i32 289, i32 295, i32 301, i32 307, i32 313, i32 319, i32 325, i32 331, i32 337, i32 343, i32 349, i32 355, i32 361, i32 367, i32 373, i32 379>
   %strided.vec2 = shufflevector <384 x i8> %wide.vec, <384 x i8> poison, <64 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92, i32 98, i32 104, i32 110, i32 116, i32 122, i32 128, i32 134, i32 140, i32 146, i32 152, i32 158, i32 164, i32 170, i32 176, i32 182, i32 188, i32 194, i32 200, i32 206, i32 212, i32 218, i32 224, i32 230, i32 236, i32 242, i32 248, i32 254, i32 260, i32 266, i32 272, i32 278, i32 284, i32 290, i32 296, i32 302, i32 308, i32 314, i32 320, i32 326, i32 332, i32 338, i32 344, i32 350, i32 356, i32 362, i32 368, i32 374, i32 380>
   %strided.vec3 = shufflevector <384 x i8> %wide.vec, <384 x i8> poison, <64 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93, i32 99, i32 105, i32 111, i32 117, i32 123, i32 129, i32 135, i32 141, i32 147, i32 153, i32 159, i32 165, i32 171, i32 177, i32 183, i32 189, i32 195, i32 201, i32 207, i32 213, i32 219, i32 225, i32 231, i32 237, i32 243, i32 249, i32 255, i32 261, i32 267, i32 273, i32 279, i32 285, i32 291, i32 297, i32 303, i32 309, i32 315, i32 321, i32 327, i32 333, i32 339, i32 345, i32 351, i32 357, i32 363, i32 369, i32 375, i32 381>
   %strided.vec4 = shufflevector <384 x i8> %wide.vec, <384 x i8> poison, <64 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94, i32 100, i32 106, i32 112, i32 118, i32 124, i32 130, i32 136, i32 142, i32 148, i32 154, i32 160, i32 166, i32 172, i32 178, i32 184, i32 190, i32 196, i32 202, i32 208, i32 214, i32 220, i32 226, i32 232, i32 238, i32 244, i32 250, i32 256, i32 262, i32 268, i32 274, i32 280, i32 286, i32 292, i32 298, i32 304, i32 310, i32 316, i32 322, i32 328, i32 334, i32 340, i32 346, i32 352, i32 358, i32 364, i32 370, i32 376, i32 382>
   %strided.vec5 = shufflevector <384 x i8> %wide.vec, <384 x i8> poison, <64 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95, i32 101, i32 107, i32 113, i32 119, i32 125, i32 131, i32 137, i32 143, i32 149, i32 155, i32 161, i32 167, i32 173, i32 179, i32 185, i32 191, i32 197, i32 203, i32 209, i32 215, i32 221, i32 227, i32 233, i32 239, i32 245, i32 251, i32 257, i32 263, i32 269, i32 275, i32 281, i32 287, i32 293, i32 299, i32 305, i32 311, i32 317, i32 323, i32 329, i32 335, i32 341, i32 347, i32 353, i32 359, i32 365, i32 371, i32 377, i32 383>
-
   store <64 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i8> %strided.vec2, ptr %out.vec2, align 64
   store <64 x i8> %strided.vec3, ptr %out.vec3, align 64
   store <64 x i8> %strided.vec4, ptr %out.vec4, align 64
   store <64 x i8> %strided.vec5, ptr %out.vec5, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
index 0e5bf8b..b7619df 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
@@ -88,7 +88,6 @@ define void @load_i8_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX-NEXT:    vpextrw $0, %xmm0, (%rax)
 ; AVX-NEXT:    retq
   %wide.vec = load <14 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <14 x i8> %wide.vec, <14 x i8> poison, <2 x i32> <i32 0, i32 7>
   %strided.vec1 = shufflevector <14 x i8> %wide.vec, <14 x i8> poison, <2 x i32> <i32 1, i32 8>
   %strided.vec2 = shufflevector <14 x i8> %wide.vec, <14 x i8> poison, <2 x i32> <i32 2, i32 9>
@@ -96,7 +95,6 @@ define void @load_i8_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <14 x i8> %wide.vec, <14 x i8> poison, <2 x i32> <i32 4, i32 11>
   %strided.vec5 = shufflevector <14 x i8> %wide.vec, <14 x i8> poison, <2 x i32> <i32 5, i32 12>
   %strided.vec6 = shufflevector <14 x i8> %wide.vec, <14 x i8> poison, <2 x i32> <i32 6, i32 13>
-
   store <2 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i8> %strided.vec2, ptr %out.vec2, align 64
@@ -104,7 +102,6 @@ define void @load_i8_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <2 x i8> %strided.vec4, ptr %out.vec4, align 64
   store <2 x i8> %strided.vec5, ptr %out.vec5, align 64
   store <2 x i8> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -255,7 +252,6 @@ define void @load_i8_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX-NEXT:    vmovd %xmm0, (%rax)
 ; AVX-NEXT:    retq
   %wide.vec = load <28 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <28 x i8> %wide.vec, <28 x i8> poison, <4 x i32> <i32 0, i32 7, i32 14, i32 21>
   %strided.vec1 = shufflevector <28 x i8> %wide.vec, <28 x i8> poison, <4 x i32> <i32 1, i32 8, i32 15, i32 22>
   %strided.vec2 = shufflevector <28 x i8> %wide.vec, <28 x i8> poison, <4 x i32> <i32 2, i32 9, i32 16, i32 23>
@@ -263,7 +259,6 @@ define void @load_i8_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <28 x i8> %wide.vec, <28 x i8> poison, <4 x i32> <i32 4, i32 11, i32 18, i32 25>
   %strided.vec5 = shufflevector <28 x i8> %wide.vec, <28 x i8> poison, <4 x i32> <i32 5, i32 12, i32 19, i32 26>
   %strided.vec6 = shufflevector <28 x i8> %wide.vec, <28 x i8> poison, <4 x i32> <i32 6, i32 13, i32 20, i32 27>
-
   store <4 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i8> %strided.vec2, ptr %out.vec2, align 64
@@ -271,7 +266,6 @@ define void @load_i8_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <4 x i8> %strided.vec4, ptr %out.vec4, align 64
   store <4 x i8> %strided.vec5, ptr %out.vec5, align 64
   store <4 x i8> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -727,7 +721,6 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <56 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <56 x i8> %wide.vec, <56 x i8> poison, <8 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49>
   %strided.vec1 = shufflevector <56 x i8> %wide.vec, <56 x i8> poison, <8 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50>
   %strided.vec2 = shufflevector <56 x i8> %wide.vec, <56 x i8> poison, <8 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51>
@@ -735,7 +728,6 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <56 x i8> %wide.vec, <56 x i8> poison, <8 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53>
   %strided.vec5 = shufflevector <56 x i8> %wide.vec, <56 x i8> poison, <8 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54>
   %strided.vec6 = shufflevector <56 x i8> %wide.vec, <56 x i8> poison, <8 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55>
-
   store <8 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i8> %strided.vec2, ptr %out.vec2, align 64
@@ -743,7 +735,6 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <8 x i8> %strided.vec4, ptr %out.vec4, align 64
   store <8 x i8> %strided.vec5, ptr %out.vec5, align 64
   store <8 x i8> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -2390,7 +2381,6 @@ define void @load_i8_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vmovdqa %xmm0, (%rax)
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <112 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <112 x i8> %wide.vec, <112 x i8> poison, <16 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49, i32 56, i32 63, i32 70, i32 77, i32 84, i32 91, i32 98, i32 105>
   %strided.vec1 = shufflevector <112 x i8> %wide.vec, <112 x i8> poison, <16 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50, i32 57, i32 64, i32 71, i32 78, i32 85, i32 92, i32 99, i32 106>
   %strided.vec2 = shufflevector <112 x i8> %wide.vec, <112 x i8> poison, <16 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51, i32 58, i32 65, i32 72, i32 79, i32 86, i32 93, i32 100, i32 107>
@@ -2398,7 +2388,6 @@ define void @load_i8_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <112 x i8> %wide.vec, <112 x i8> poison, <16 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53, i32 60, i32 67, i32 74, i32 81, i32 88, i32 95, i32 102, i32 109>
   %strided.vec5 = shufflevector <112 x i8> %wide.vec, <112 x i8> poison, <16 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54, i32 61, i32 68, i32 75, i32 82, i32 89, i32 96, i32 103, i32 110>
   %strided.vec6 = shufflevector <112 x i8> %wide.vec, <112 x i8> poison, <16 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55, i32 62, i32 69, i32 76, i32 83, i32 90, i32 97, i32 104, i32 111>
-
   store <16 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i8> %strided.vec2, ptr %out.vec2, align 64
@@ -2406,7 +2395,6 @@ define void @load_i8_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <16 x i8> %strided.vec4, ptr %out.vec4, align 64
   store <16 x i8> %strided.vec5, ptr %out.vec5, align 64
   store <16 x i8> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -5213,7 +5201,6 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <224 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <224 x i8> %wide.vec, <224 x i8> poison, <32 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49, i32 56, i32 63, i32 70, i32 77, i32 84, i32 91, i32 98, i32 105, i32 112, i32 119, i32 126, i32 133, i32 140, i32 147, i32 154, i32 161, i32 168, i32 175, i32 182, i32 189, i32 196, i32 203, i32 210, i32 217>
   %strided.vec1 = shufflevector <224 x i8> %wide.vec, <224 x i8> poison, <32 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50, i32 57, i32 64, i32 71, i32 78, i32 85, i32 92, i32 99, i32 106, i32 113, i32 120, i32 127, i32 134, i32 141, i32 148, i32 155, i32 162, i32 169, i32 176, i32 183, i32 190, i32 197, i32 204, i32 211, i32 218>
   %strided.vec2 = shufflevector <224 x i8> %wide.vec, <224 x i8> poison, <32 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51, i32 58, i32 65, i32 72, i32 79, i32 86, i32 93, i32 100, i32 107, i32 114, i32 121, i32 128, i32 135, i32 142, i32 149, i32 156, i32 163, i32 170, i32 177, i32 184, i32 191, i32 198, i32 205, i32 212, i32 219>
@@ -5221,7 +5208,6 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <224 x i8> %wide.vec, <224 x i8> poison, <32 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53, i32 60, i32 67, i32 74, i32 81, i32 88, i32 95, i32 102, i32 109, i32 116, i32 123, i32 130, i32 137, i32 144, i32 151, i32 158, i32 165, i32 172, i32 179, i32 186, i32 193, i32 200, i32 207, i32 214, i32 221>
   %strided.vec5 = shufflevector <224 x i8> %wide.vec, <224 x i8> poison, <32 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54, i32 61, i32 68, i32 75, i32 82, i32 89, i32 96, i32 103, i32 110, i32 117, i32 124, i32 131, i32 138, i32 145, i32 152, i32 159, i32 166, i32 173, i32 180, i32 187, i32 194, i32 201, i32 208, i32 215, i32 222>
   %strided.vec6 = shufflevector <224 x i8> %wide.vec, <224 x i8> poison, <32 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55, i32 62, i32 69, i32 76, i32 83, i32 90, i32 97, i32 104, i32 111, i32 118, i32 125, i32 132, i32 139, i32 146, i32 153, i32 160, i32 167, i32 174, i32 181, i32 188, i32 195, i32 202, i32 209, i32 216, i32 223>
-
   store <32 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i8> %strided.vec2, ptr %out.vec2, align 64
@@ -5229,7 +5215,6 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <32 x i8> %strided.vec4, ptr %out.vec4, align 64
   store <32 x i8> %strided.vec5, ptr %out.vec5, align 64
   store <32 x i8> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 
@@ -12763,7 +12748,6 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512DQBW-FAST-NEXT:    vzeroupper
 ; AVX512DQBW-FAST-NEXT:    retq
   %wide.vec = load <448 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <448 x i8> %wide.vec, <448 x i8> poison, <64 x i32> <i32 0, i32 7, i32 14, i32 21, i32 28, i32 35, i32 42, i32 49, i32 56, i32 63, i32 70, i32 77, i32 84, i32 91, i32 98, i32 105, i32 112, i32 119, i32 126, i32 133, i32 140, i32 147, i32 154, i32 161, i32 168, i32 175, i32 182, i32 189, i32 196, i32 203, i32 210, i32 217, i32 224, i32 231, i32 238, i32 245, i32 252, i32 259, i32 266, i32 273, i32 280, i32 287, i32 294, i32 301, i32 308, i32 315, i32 322, i32 329, i32 336, i32 343, i32 350, i32 357, i32 364, i32 371, i32 378, i32 385, i32 392, i32 399, i32 406, i32 413, i32 420, i32 427, i32 434, i32 441>
   %strided.vec1 = shufflevector <448 x i8> %wide.vec, <448 x i8> poison, <64 x i32> <i32 1, i32 8, i32 15, i32 22, i32 29, i32 36, i32 43, i32 50, i32 57, i32 64, i32 71, i32 78, i32 85, i32 92, i32 99, i32 106, i32 113, i32 120, i32 127, i32 134, i32 141, i32 148, i32 155, i32 162, i32 169, i32 176, i32 183, i32 190, i32 197, i32 204, i32 211, i32 218, i32 225, i32 232, i32 239, i32 246, i32 253, i32 260, i32 267, i32 274, i32 281, i32 288, i32 295, i32 302, i32 309, i32 316, i32 323, i32 330, i32 337, i32 344, i32 351, i32 358, i32 365, i32 372, i32 379, i32 386, i32 393, i32 400, i32 407, i32 414, i32 421, i32 428, i32 435, i32 442>
   %strided.vec2 = shufflevector <448 x i8> %wide.vec, <448 x i8> poison, <64 x i32> <i32 2, i32 9, i32 16, i32 23, i32 30, i32 37, i32 44, i32 51, i32 58, i32 65, i32 72, i32 79, i32 86, i32 93, i32 100, i32 107, i32 114, i32 121, i32 128, i32 135, i32 142, i32 149, i32 156, i32 163, i32 170, i32 177, i32 184, i32 191, i32 198, i32 205, i32 212, i32 219, i32 226, i32 233, i32 240, i32 247, i32 254, i32 261, i32 268, i32 275, i32 282, i32 289, i32 296, i32 303, i32 310, i32 317, i32 324, i32 331, i32 338, i32 345, i32 352, i32 359, i32 366, i32 373, i32 380, i32 387, i32 394, i32 401, i32 408, i32 415, i32 422, i32 429, i32 436, i32 443>
@@ -12771,7 +12755,6 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec4 = shufflevector <448 x i8> %wide.vec, <448 x i8> poison, <64 x i32> <i32 4, i32 11, i32 18, i32 25, i32 32, i32 39, i32 46, i32 53, i32 60, i32 67, i32 74, i32 81, i32 88, i32 95, i32 102, i32 109, i32 116, i32 123, i32 130, i32 137, i32 144, i32 151, i32 158, i32 165, i32 172, i32 179, i32 186, i32 193, i32 200, i32 207, i32 214, i32 221, i32 228, i32 235, i32 242, i32 249, i32 256, i32 263, i32 270, i32 277, i32 284, i32 291, i32 298, i32 305, i32 312, i32 319, i32 326, i32 333, i32 340, i32 347, i32 354, i32 361, i32 368, i32 375, i32 382, i32 389, i32 396, i32 403, i32 410, i32 417, i32 424, i32 431, i32 438, i32 445>
   %strided.vec5 = shufflevector <448 x i8> %wide.vec, <448 x i8> poison, <64 x i32> <i32 5, i32 12, i32 19, i32 26, i32 33, i32 40, i32 47, i32 54, i32 61, i32 68, i32 75, i32 82, i32 89, i32 96, i32 103, i32 110, i32 117, i32 124, i32 131, i32 138, i32 145, i32 152, i32 159, i32 166, i32 173, i32 180, i32 187, i32 194, i32 201, i32 208, i32 215, i32 222, i32 229, i32 236, i32 243, i32 250, i32 257, i32 264, i32 271, i32 278, i32 285, i32 292, i32 299, i32 306, i32 313, i32 320, i32 327, i32 334, i32 341, i32 348, i32 355, i32 362, i32 369, i32 376, i32 383, i32 390, i32 397, i32 404, i32 411, i32 418, i32 425, i32 432, i32 439, i32 446>
   %strided.vec6 = shufflevector <448 x i8> %wide.vec, <448 x i8> poison, <64 x i32> <i32 6, i32 13, i32 20, i32 27, i32 34, i32 41, i32 48, i32 55, i32 62, i32 69, i32 76, i32 83, i32 90, i32 97, i32 104, i32 111, i32 118, i32 125, i32 132, i32 139, i32 146, i32 153, i32 160, i32 167, i32 174, i32 181, i32 188, i32 195, i32 202, i32 209, i32 216, i32 223, i32 230, i32 237, i32 244, i32 251, i32 258, i32 265, i32 272, i32 279, i32 286, i32 293, i32 300, i32 307, i32 314, i32 321, i32 328, i32 335, i32 342, i32 349, i32 356, i32 363, i32 370, i32 377, i32 384, i32 391, i32 398, i32 405, i32 412, i32 419, i32 426, i32 433, i32 440, i32 447>
-
   store <64 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i8> %strided.vec2, ptr %out.vec2, align 64
@@ -12779,7 +12762,6 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <64 x i8> %strided.vec4, ptr %out.vec4, align 64
   store <64 x i8> %strided.vec5, ptr %out.vec5, align 64
   store <64 x i8> %strided.vec6, ptr %out.vec6, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll
index d7d7f99..b21d8eb 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll
@@ -117,7 +117,6 @@ define void @load_i8_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vpextrw $0, %xmm7, (%rax)
 ; AVX512-NEXT:    retq
   %wide.vec = load <16 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <16 x i8> %wide.vec, <16 x i8> poison, <2 x i32> <i32 0, i32 8>
   %strided.vec1 = shufflevector <16 x i8> %wide.vec, <16 x i8> poison, <2 x i32> <i32 1, i32 9>
   %strided.vec2 = shufflevector <16 x i8> %wide.vec, <16 x i8> poison, <2 x i32> <i32 2, i32 10>
@@ -126,7 +125,6 @@ define void @load_i8_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <16 x i8> %wide.vec, <16 x i8> poison, <2 x i32> <i32 5, i32 13>
   %strided.vec6 = shufflevector <16 x i8> %wide.vec, <16 x i8> poison, <2 x i32> <i32 6, i32 14>
   %strided.vec7 = shufflevector <16 x i8> %wide.vec, <16 x i8> poison, <2 x i32> <i32 7, i32 15>
-
   store <2 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <2 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <2 x i8> %strided.vec2, ptr %out.vec2, align 64
@@ -135,7 +133,6 @@ define void @load_i8_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <2 x i8> %strided.vec5, ptr %out.vec5, align 64
   store <2 x i8> %strided.vec6, ptr %out.vec6, align 64
   store <2 x i8> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -334,7 +331,6 @@ define void @load_i8_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %wide.vec = load <32 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <32 x i8> %wide.vec, <32 x i8> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
   %strided.vec1 = shufflevector <32 x i8> %wide.vec, <32 x i8> poison, <4 x i32> <i32 1, i32 9, i32 17, i32 25>
   %strided.vec2 = shufflevector <32 x i8> %wide.vec, <32 x i8> poison, <4 x i32> <i32 2, i32 10, i32 18, i32 26>
@@ -343,7 +339,6 @@ define void @load_i8_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <32 x i8> %wide.vec, <32 x i8> poison, <4 x i32> <i32 5, i32 13, i32 21, i32 29>
   %strided.vec6 = shufflevector <32 x i8> %wide.vec, <32 x i8> poison, <4 x i32> <i32 6, i32 14, i32 22, i32 30>
   %strided.vec7 = shufflevector <32 x i8> %wide.vec, <32 x i8> poison, <4 x i32> <i32 7, i32 15, i32 23, i32 31>
-
   store <4 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <4 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <4 x i8> %strided.vec2, ptr %out.vec2, align 64
@@ -352,7 +347,6 @@ define void @load_i8_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <4 x i8> %strided.vec5, ptr %out.vec5, align 64
   store <4 x i8> %strided.vec6, ptr %out.vec6, align 64
   store <4 x i8> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -917,7 +911,6 @@ define void @load_i8_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %wide.vec = load <64 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <8 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56>
   %strided.vec1 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <8 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57>
   %strided.vec2 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <8 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58>
@@ -926,7 +919,6 @@ define void @load_i8_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <8 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61>
   %strided.vec6 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <8 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62>
   %strided.vec7 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <8 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63>
-
   store <8 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <8 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <8 x i8> %strided.vec2, ptr %out.vec2, align 64
@@ -935,7 +927,6 @@ define void @load_i8_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <8 x i8> %strided.vec5, ptr %out.vec5, align 64
   store <8 x i8> %strided.vec6, ptr %out.vec6, align 64
   store <8 x i8> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -2826,7 +2817,6 @@ define void @load_i8_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    popq %rbx
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <128 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <128 x i8> %wide.vec, <128 x i8> poison, <16 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120>
   %strided.vec1 = shufflevector <128 x i8> %wide.vec, <128 x i8> poison, <16 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121>
   %strided.vec2 = shufflevector <128 x i8> %wide.vec, <128 x i8> poison, <16 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122>
@@ -2835,7 +2825,6 @@ define void @load_i8_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <128 x i8> %wide.vec, <128 x i8> poison, <16 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125>
   %strided.vec6 = shufflevector <128 x i8> %wide.vec, <128 x i8> poison, <16 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126>
   %strided.vec7 = shufflevector <128 x i8> %wide.vec, <128 x i8> poison, <16 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127>
-
   store <16 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <16 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <16 x i8> %strided.vec2, ptr %out.vec2, align 64
@@ -2844,7 +2833,6 @@ define void @load_i8_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <16 x i8> %strided.vec5, ptr %out.vec5, align 64
   store <16 x i8> %strided.vec6, ptr %out.vec6, align 64
   store <16 x i8> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -6780,7 +6768,6 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <256 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <256 x i8> %wide.vec, <256 x i8> poison, <32 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120, i32 128, i32 136, i32 144, i32 152, i32 160, i32 168, i32 176, i32 184, i32 192, i32 200, i32 208, i32 216, i32 224, i32 232, i32 240, i32 248>
   %strided.vec1 = shufflevector <256 x i8> %wide.vec, <256 x i8> poison, <32 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121, i32 129, i32 137, i32 145, i32 153, i32 161, i32 169, i32 177, i32 185, i32 193, i32 201, i32 209, i32 217, i32 225, i32 233, i32 241, i32 249>
   %strided.vec2 = shufflevector <256 x i8> %wide.vec, <256 x i8> poison, <32 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122, i32 130, i32 138, i32 146, i32 154, i32 162, i32 170, i32 178, i32 186, i32 194, i32 202, i32 210, i32 218, i32 226, i32 234, i32 242, i32 250>
@@ -6789,7 +6776,6 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <256 x i8> %wide.vec, <256 x i8> poison, <32 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125, i32 133, i32 141, i32 149, i32 157, i32 165, i32 173, i32 181, i32 189, i32 197, i32 205, i32 213, i32 221, i32 229, i32 237, i32 245, i32 253>
   %strided.vec6 = shufflevector <256 x i8> %wide.vec, <256 x i8> poison, <32 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126, i32 134, i32 142, i32 150, i32 158, i32 166, i32 174, i32 182, i32 190, i32 198, i32 206, i32 214, i32 222, i32 230, i32 238, i32 246, i32 254>
   %strided.vec7 = shufflevector <256 x i8> %wide.vec, <256 x i8> poison, <32 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127, i32 135, i32 143, i32 151, i32 159, i32 167, i32 175, i32 183, i32 191, i32 199, i32 207, i32 215, i32 223, i32 231, i32 239, i32 247, i32 255>
-
   store <32 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <32 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <32 x i8> %strided.vec2, ptr %out.vec2, align 64
@@ -6798,7 +6784,6 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <32 x i8> %strided.vec5, ptr %out.vec5, align 64
   store <32 x i8> %strided.vec6, ptr %out.vec6, align 64
   store <32 x i8> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 
@@ -15070,7 +15055,6 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX512BW-FAST-NEXT:    vzeroupper
 ; AVX512BW-FAST-NEXT:    retq
   %wide.vec = load <512 x i8>, ptr %in.vec, align 64
-
   %strided.vec0 = shufflevector <512 x i8> %wide.vec, <512 x i8> poison, <64 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120, i32 128, i32 136, i32 144, i32 152, i32 160, i32 168, i32 176, i32 184, i32 192, i32 200, i32 208, i32 216, i32 224, i32 232, i32 240, i32 248, i32 256, i32 264, i32 272, i32 280, i32 288, i32 296, i32 304, i32 312, i32 320, i32 328, i32 336, i32 344, i32 352, i32 360, i32 368, i32 376, i32 384, i32 392, i32 400, i32 408, i32 416, i32 424, i32 432, i32 440, i32 448, i32 456, i32 464, i32 472, i32 480, i32 488, i32 496, i32 504>
   %strided.vec1 = shufflevector <512 x i8> %wide.vec, <512 x i8> poison, <64 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121, i32 129, i32 137, i32 145, i32 153, i32 161, i32 169, i32 177, i32 185, i32 193, i32 201, i32 209, i32 217, i32 225, i32 233, i32 241, i32 249, i32 257, i32 265, i32 273, i32 281, i32 289, i32 297, i32 305, i32 313, i32 321, i32 329, i32 337, i32 345, i32 353, i32 361, i32 369, i32 377, i32 385, i32 393, i32 401, i32 409, i32 417, i32 425, i32 433, i32 441, i32 449, i32 457, i32 465, i32 473, i32 481, i32 489, i32 497, i32 505>
   %strided.vec2 = shufflevector <512 x i8> %wide.vec, <512 x i8> poison, <64 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122, i32 130, i32 138, i32 146, i32 154, i32 162, i32 170, i32 178, i32 186, i32 194, i32 202, i32 210, i32 218, i32 226, i32 234, i32 242, i32 250, i32 258, i32 266, i32 274, i32 282, i32 290, i32 298, i32 306, i32 314, i32 322, i32 330, i32 338, i32 346, i32 354, i32 362, i32 370, i32 378, i32 386, i32 394, i32 402, i32 410, i32 418, i32 426, i32 434, i32 442, i32 450, i32 458, i32 466, i32 474, i32 482, i32 490, i32 498, i32 506>
@@ -15079,7 +15063,6 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   %strided.vec5 = shufflevector <512 x i8> %wide.vec, <512 x i8> poison, <64 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125, i32 133, i32 141, i32 149, i32 157, i32 165, i32 173, i32 181, i32 189, i32 197, i32 205, i32 213, i32 221, i32 229, i32 237, i32 245, i32 253, i32 261, i32 269, i32 277, i32 285, i32 293, i32 301, i32 309, i32 317, i32 325, i32 333, i32 341, i32 349, i32 357, i32 365, i32 373, i32 381, i32 389, i32 397, i32 405, i32 413, i32 421, i32 429, i32 437, i32 445, i32 453, i32 461, i32 469, i32 477, i32 485, i32 493, i32 501, i32 509>
   %strided.vec6 = shufflevector <512 x i8> %wide.vec, <512 x i8> poison, <64 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126, i32 134, i32 142, i32 150, i32 158, i32 166, i32 174, i32 182, i32 190, i32 198, i32 206, i32 214, i32 222, i32 230, i32 238, i32 246, i32 254, i32 262, i32 270, i32 278, i32 286, i32 294, i32 302, i32 310, i32 318, i32 326, i32 334, i32 342, i32 350, i32 358, i32 366, i32 374, i32 382, i32 390, i32 398, i32 406, i32 414, i32 422, i32 430, i32 438, i32 446, i32 454, i32 462, i32 470, i32 478, i32 486, i32 494, i32 502, i32 510>
   %strided.vec7 = shufflevector <512 x i8> %wide.vec, <512 x i8> poison, <64 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127, i32 135, i32 143, i32 151, i32 159, i32 167, i32 175, i32 183, i32 191, i32 199, i32 207, i32 215, i32 223, i32 231, i32 239, i32 247, i32 255, i32 263, i32 271, i32 279, i32 287, i32 295, i32 303, i32 311, i32 319, i32 327, i32 335, i32 343, i32 351, i32 359, i32 367, i32 375, i32 383, i32 391, i32 399, i32 407, i32 415, i32 423, i32 431, i32 439, i32 447, i32 455, i32 463, i32 471, i32 479, i32 487, i32 495, i32 503, i32 511>
-
   store <64 x i8> %strided.vec0, ptr %out.vec0, align 64
   store <64 x i8> %strided.vec1, ptr %out.vec1, align 64
   store <64 x i8> %strided.vec2, ptr %out.vec2, align 64
@@ -15088,7 +15071,6 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
   store <64 x i8> %strided.vec5, ptr %out.vec5, align 64
   store <64 x i8> %strided.vec6, ptr %out.vec6, align 64
   store <64 x i8> %strided.vec7, ptr %out.vec7, align 64
-
   ret void
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: