[TLI] SimplifyMultipleUseDemandedBits - remove insert_subvector(undef, x, 0) fold

SimplifyMultipleUseDemandedBits shouldn't be creating general nodes on the fly, it should mainly just peek through them (although we do currently allow creation of new bitcasts and constant folding).

This is mostly a win - by avoiding new nodes we avoid a lot of hasOneUse limitations inside x86 shuffle combining - the main regressions I've noticed are where we've ended up with multiple insert_subvector(undef, x, 0) nodes, widening x to different vector widths - that should hopefully be improved when we remove the last of the vector widening from combineX86ShufflesRecursively for Issue #45319

diff --git a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index db26fd2..9c5743f 100644 (file)
--- a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -855,15 +855,6 @@ SDValue TargetLowering::SimplifyMultipleUseDemandedBits(
     // If we don't demand the inserted subvector, return the base vector.
     if (DemandedSubElts == 0)
       return Vec;
-    // If this simply widens the lowest subvector, see if we can do it earlier.
-    // TODO: REMOVE ME - SimplifyMultipleUseDemandedBits shouldn't be creating
-    // general nodes like this.
-    if (Idx == 0 && Vec.isUndef()) {
-      if (SDValue NewSub = SimplifyMultipleUseDemandedBits(
-              Sub, DemandedBits, DemandedSubElts, DAG, Depth + 1))
-        return DAG.getNode(Op.getOpcode(), SDLoc(Op), Op.getValueType(),
-                           Op.getOperand(0), NewSub, Op.getOperand(2));
-    }
     break;
   }
   case ISD::VECTOR_SHUFFLE: {

diff --git a/llvm/test/CodeGen/AArch64/vecreduce-and-legalization.ll b/llvm/test/CodeGen/AArch64/vecreduce-and-legalization.ll
index 95a524b..d7a63af 100644 (file)
--- a/llvm/test/CodeGen/AArch64/vecreduce-and-legalization.ll
+++ b/llvm/test/CodeGen/AArch64/vecreduce-and-legalization.ll
@@ -97,28 +97,28 @@ define i8 @test_v9i8(<9 x i8> %a) nounwind {
 ; CHECK-LABEL: test_v9i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov w8, #-1
-; CHECK-NEXT:    umov w9, v0.b[5]
+; CHECK-NEXT:    umov w14, v0.b[6]
 ; CHECK-NEXT:    mov v1.16b, v0.16b
-; CHECK-NEXT:    umov w10, v0.b[6]
-; CHECK-NEXT:    umov w15, v0.b[7]
 ; CHECK-NEXT:    mov v1.b[9], w8
 ; CHECK-NEXT:    mov v1.b[10], w8
 ; CHECK-NEXT:    mov v1.b[11], w8
 ; CHECK-NEXT:    mov v1.b[13], w8
 ; CHECK-NEXT:    umov w8, v0.b[4]
 ; CHECK-NEXT:    ext v1.16b, v1.16b, v1.16b, #8
-; CHECK-NEXT:    and w8, w8, w9
-; CHECK-NEXT:    and w8, w8, w10
-; CHECK-NEXT:    and w8, w8, w15
 ; CHECK-NEXT:    and v1.8b, v0.8b, v1.8b
-; CHECK-NEXT:    umov w11, v1.b[1]
-; CHECK-NEXT:    umov w12, v1.b[0]
-; CHECK-NEXT:    umov w13, v1.b[2]
-; CHECK-NEXT:    umov w14, v1.b[3]
-; CHECK-NEXT:    and w9, w12, w11
-; CHECK-NEXT:    and w11, w13, w14
+; CHECK-NEXT:    umov w9, v1.b[1]
+; CHECK-NEXT:    umov w10, v1.b[0]
+; CHECK-NEXT:    umov w11, v1.b[2]
+; CHECK-NEXT:    umov w12, v1.b[3]
+; CHECK-NEXT:    umov w13, v1.b[5]
+; CHECK-NEXT:    and w9, w10, w9
+; CHECK-NEXT:    umov w10, v0.b[7]
+; CHECK-NEXT:    and w11, w11, w12
+; CHECK-NEXT:    and w8, w8, w13
 ; CHECK-NEXT:    and w9, w9, w11
-; CHECK-NEXT:    and w0, w9, w8
+; CHECK-NEXT:    and w8, w8, w14
+; CHECK-NEXT:    and w8, w9, w8
+; CHECK-NEXT:    and w0, w8, w10
 ; CHECK-NEXT:    ret
   %b = call i8 @llvm.vector.reduce.and.v9i8(<9 x i8> %a)
   ret i8 %b

diff --git a/llvm/test/CodeGen/X86/horizontal-sum.ll b/llvm/test/CodeGen/X86/horizontal-sum.ll
index 8dab14a..c1011a6 100644 (file)
--- a/llvm/test/CodeGen/X86/horizontal-sum.ll
+++ b/llvm/test/CodeGen/X86/horizontal-sum.ll
@@ -229,13 +229,13 @@ define <8 x float> @pair_sum_v8f32_v4f32(<4 x float> %0, <4 x float> %1, <4 x fl
 ; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
 ; AVX1-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
 ; AVX1-SLOW-NEXT:    vaddps %xmm1, %xmm3, %xmm1
-; AVX1-SLOW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
-; AVX1-SLOW-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-SLOW-NEXT:    vmovddup {{.*#+}} xmm1 = xmm1[0,0]
+; AVX1-SLOW-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-SLOW-NEXT:    vhaddps %xmm7, %xmm6, %xmm2
 ; AVX1-SLOW-NEXT:    vhaddps %xmm2, %xmm2, %xmm2
-; AVX1-SLOW-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX1-SLOW-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[2]
+; AVX1-SLOW-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-SLOW-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[2]
 ; AVX1-SLOW-NEXT:    retq
 ;
 ; AVX1-FAST-LABEL: pair_sum_v8f32_v4f32:
@@ -250,13 +250,13 @@ define <8 x float> @pair_sum_v8f32_v4f32(<4 x float> %0, <4 x float> %1, <4 x fl
 ; AVX1-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
 ; AVX1-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
 ; AVX1-FAST-NEXT:    vaddps %xmm1, %xmm3, %xmm1
-; AVX1-FAST-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
-; AVX1-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-FAST-NEXT:    vmovddup {{.*#+}} xmm1 = xmm1[0,0]
+; AVX1-FAST-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-FAST-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-FAST-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-FAST-NEXT:    vhaddps %xmm7, %xmm6, %xmm2
 ; AVX1-FAST-NEXT:    vhaddps %xmm0, %xmm2, %xmm2
-; AVX1-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX1-FAST-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[2]
+; AVX1-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-FAST-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[2]
 ; AVX1-FAST-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: pair_sum_v8f32_v4f32:
@@ -273,13 +273,13 @@ define <8 x float> @pair_sum_v8f32_v4f32(<4 x float> %0, <4 x float> %1, <4 x fl
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
 ; AVX2-SLOW-NEXT:    vaddps %xmm1, %xmm3, %xmm1
-; AVX2-SLOW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vmovddup {{.*#+}} xmm1 = xmm1[0,0]
+; AVX2-SLOW-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX2-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-SLOW-NEXT:    vhaddps %xmm7, %xmm6, %xmm2
 ; AVX2-SLOW-NEXT:    vhaddps %xmm2, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[2]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: pair_sum_v8f32_v4f32:
@@ -294,13 +294,13 @@ define <8 x float> @pair_sum_v8f32_v4f32(<4 x float> %0, <4 x float> %1, <4 x fl
 ; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
 ; AVX2-FAST-NEXT:    vaddps %xmm1, %xmm3, %xmm1
-; AVX2-FAST-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
-; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX2-FAST-NEXT:    vmovddup {{.*#+}} xmm1 = xmm1[0,0]
+; AVX2-FAST-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX2-FAST-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-FAST-NEXT:    vhaddps %xmm7, %xmm6, %xmm2
 ; AVX2-FAST-NEXT:    vhaddps %xmm0, %xmm2, %xmm2
-; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX2-FAST-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[2]
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[2]
 ; AVX2-FAST-NEXT:    retq
   %9 = shufflevector <4 x float> %0, <4 x float> poison, <2 x i32> <i32 0, i32 2>
   %10 = shufflevector <4 x float> %0, <4 x float> poison, <2 x i32> <i32 1, i32 3>
@@ -399,13 +399,13 @@ define <8 x i32> @pair_sum_v8i32_v4i32(<4 x i32> %0, <4 x i32> %1, <4 x i32> %2,
 ; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
 ; AVX1-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
 ; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm3, %xmm1
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[2,3,2,3]
-; AVX1-SLOW-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-SLOW-NEXT:    vphaddd %xmm7, %xmm6, %xmm2
 ; AVX1-SLOW-NEXT:    vphaddd %xmm2, %xmm2, %xmm2
-; AVX1-SLOW-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX1-SLOW-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[2]
+; AVX1-SLOW-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-SLOW-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[2]
 ; AVX1-SLOW-NEXT:    retq
 ;
 ; AVX1-FAST-LABEL: pair_sum_v8i32_v4i32:
@@ -422,13 +422,13 @@ define <8 x i32> @pair_sum_v8i32_v4i32(<4 x i32> %0, <4 x i32> %1, <4 x i32> %2,
 ; AVX1-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
 ; AVX1-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
 ; AVX1-FAST-NEXT:    vpaddd %xmm1, %xmm3, %xmm1
-; AVX1-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[2,3,2,3]
-; AVX1-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-FAST-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-FAST-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-FAST-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-FAST-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-FAST-NEXT:    vphaddd %xmm7, %xmm6, %xmm2
 ; AVX1-FAST-NEXT:    vphaddd %xmm0, %xmm2, %xmm2
-; AVX1-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX1-FAST-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[2]
+; AVX1-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-FAST-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[2]
 ; AVX1-FAST-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: pair_sum_v8i32_v4i32:

diff --git a/llvm/test/CodeGen/X86/oddshuffles.ll b/llvm/test/CodeGen/X86/oddshuffles.ll
index f8e9bf3..ac028c1 100644 (file)
--- a/llvm/test/CodeGen/X86/oddshuffles.ll
+++ b/llvm/test/CodeGen/X86/oddshuffles.ll
@@ -2253,7 +2253,8 @@ define <16 x i32> @splat_v3i32(ptr %ptr) {
 ;
 ; AVX2-FAST-LABEL: splat_v3i32:
 ; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
+; AVX2-FAST-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
+; AVX2-FAST-NEXT:    vpinsrd $2, 8(%rdi), %xmm0, %xmm1
 ; AVX2-FAST-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0,1,2,3],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero

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 b644b92..487e744 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
@@ -417,14 +417,14 @@ define void @load_i32_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm4[0,1],ymm3[2,3],ymm4[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm7
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0],xmm7[1],xmm5[2],xmm7[3]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0],xmm6[1],xmm5[2],xmm6[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%rdi), %ymm5, %ymm7
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm6
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm1[4],ymm9[5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm9[4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%rdi), %ymm7, %ymm7
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3],ymm5[4,5,6],ymm7[7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm2[2,3,0,1]
@@ -834,26 +834,26 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%rdi), %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4],ymm3[5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[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:    vblendps {{.*#+}} ymm0 = ymm12[0,1],ymm4[2,3],ymm12[4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%rdi), %ymm1, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 256(%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:    vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3,4,5],ymm11[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm2[4],ymm4[5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%rdi), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3],ymm1[4,5,6],ymm3[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm6[2,3],ymm5[4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm6[2,3,0,1]
@@ -1706,93 +1706,92 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX1-ONLY-LABEL: load_i32_stride5_vf32:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $984, %rsp # imm = 0x3D8
-; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %ymm9
-; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %ymm4
-; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    subq $920, %rsp # imm = 0x398
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm3
 ; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %ymm5
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm13
-; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %ymm8
+; 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), %ymm8
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %ymm9
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm0
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps %ymm2, %ymm7
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, %ymm12
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%rdi), %ymm0, %ymm1
 ; 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, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm11[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[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 = ymm5[0,1],ymm3[2,3],ymm5[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[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],ymm4[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:    vmovaps 32(%rdi), %ymm2
-; 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 (%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 608(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
 ; 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:    vmovaps %ymm2, %ymm15
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm3[2,3],ymm1[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, %ymm15
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; 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), %ymm14
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm11
-; 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, %ymm12
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm12[0,1,2,3,4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[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:    vmovaps 96(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm10
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[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, %xmm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3]
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %ymm5
+; 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), %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4],ymm3[5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%rdi), %ymm2, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3],ymm0[4,5,6],ymm2[7]
+; 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],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3],ymm2[4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm8[2,3,0,1]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm13[1,3],ymm0[6,5],ymm13[5,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm13[0,1],ymm8[2,3],ymm13[4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm9[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm8[1,3],ymm0[6,5],ymm8[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0,1],ymm9[2,3],ymm8[4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[3,0],ymm0[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 160(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-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 = xmm6[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 {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
@@ -1801,30 +1800,30 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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:    vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm9[1,3],ymm0[6,5],ymm9[5,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm4[2,3],ymm9[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm7[1,3],ymm0[6,5],ymm7[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm4[2,3],ymm7[4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[3,0],ymm0[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 480(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-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 = xmm4[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 {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 624(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[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 %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm11[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm5[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm14[1,3],ymm0[6,5],ymm14[5,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm14[0,1],ymm11[2,3],ymm14[4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm14[0,1],ymm5[2,3],ymm14[4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[3,0],ymm0[6,4],ymm2[7,4]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-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 = xmm3[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm4[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 {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
@@ -1833,8 +1832,8 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; 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:    vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm5[1,3],ymm0[6,5],ymm5[5,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm1[2,3],ymm5[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm3[1,3],ymm0[6,5],ymm3[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[3,0],ymm0[6,4],ymm2[7,4]
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1847,25 +1846,24 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX1-ONLY-NEXT:    vbroadcastss 464(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[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:    vshufps {{.*#+}} ymm0 = ymm6[3,0],ymm13[2,0],ymm6[7,4],ymm13[6,4]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm13[2,1],ymm0[6,4],ymm13[6,5]
-; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = ymm7[0,1,2,3],mem[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[3,0],ymm8[2,0],ymm11[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:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm12[0,1,2,3],mem[4,5],ymm12[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 {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm6
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm6[2,3,0,1]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm6[1,0],ymm10[0,0],ymm6[5,4],ymm10[4,4]
-; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm5[1,0],ymm1[0,0],ymm5[5,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, %ymm11
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-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 = ymm0[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:    vmovaps %ymm9, %ymm11
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm9[2,0],ymm0[7,4],ymm9[6,4]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm9[2,1],ymm0[6,4],ymm9[6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[3,0],ymm7[2,0],ymm13[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 $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
@@ -1873,31 +1871,29 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm8
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm8[2,3,0,1]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm8[1,0],ymm7[0,0],ymm8[5,4],ymm7[4,4]
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm6[1,0],ymm12[0,0],ymm6[5,4],ymm12[4,4]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm12[3,0],ymm14[2,0],ymm12[7,4],ymm14[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm10[3,0],ymm14[2,0],ymm10[7,4],ymm14[6,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm14[2,1],ymm0[6,4],ymm14[6,5]
-; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm4 = mem[0,1,2,3],ymm15[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = ymm15[0,1,2,3],mem[4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm4 = xmm4[1,0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm12
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm12[2,3,0,1]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm12[1,0],ymm3[0,0],ymm12[5,4],ymm3[4,4]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm9[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm9[1,0],ymm13[0,0],ymm9[5,4],ymm13[4,4]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[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],ymm5[2,0],ymm0[7,4],ymm5[6,4]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm5[2,1],ymm0[6,4],ymm5[6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm3[2,0],ymm0[7,4],ymm3[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm3[2,1],ymm0[6,4],ymm3[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, %ymm4 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm4 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
@@ -1906,8 +1902,8 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm4 = xmm4[1,0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm4
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm4[2,3,0,1]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[1,0],ymm9[0,0],ymm4[5,4],ymm9[4,4]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm4[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[1,0],ymm10[0,0],ymm4[5,4],ymm10[4,4]
 ; 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
@@ -1915,11 +1911,11 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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],ymm13[3,0],ymm1[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 {{.*#+}} ymm15 = ymm1[0,0],ymm8[3,0],ymm1[4,4],ymm8[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm8[2,2],ymm15[6,4],ymm8[6,6]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm6[2,0],ymm10[1,0],ymm6[6,4],ymm10[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm5[2,0],ymm11[1,0],ymm5[6,4],ymm11[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
@@ -1927,23 +1923,23 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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 {{.*#+}} ymm15 = ymm2[0,0],ymm7[3,0],ymm2[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[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm8[2,0],ymm7[1,0],ymm8[6,4],ymm7[5,4]
+; 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), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm10[0,0],ymm14[3,0],ymm10[4,4],ymm14[7,4]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm11[0,0],ymm14[3,0],ymm11[4,4],ymm14[7,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm14[2,2],ymm15[6,4],ymm14[6,6]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm12[2,0],ymm3[1,0],ymm12[6,4],ymm3[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm9[2,0],ymm13[1,0],ymm9[6,4],ymm13[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
@@ -1951,81 +1947,79 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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],ymm5[3,0],ymm1[4,4],ymm5[7,4]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm5[2,2],ymm15[6,4],ymm5[6,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm3[3,0],ymm1[4,4],ymm3[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm3[2,2],ymm15[6,4],ymm3[6,6]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[2,0],ymm9[1,0],ymm4[6,4],ymm9[5,4]
+; 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 {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    vblendps $32, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm7 = ymm7[0,1,2,3,4],mem[5],ymm7[6,7]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm15 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps $32, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm8 = ymm8[0,1,2,3,4],mem[5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = mem[0,1,2,3],ymm15[4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],mem[1],xmm15[2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm15[0,1,2],ymm7[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm3 = mem[0,1,2,3,4,5,6],ymm6[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm2[5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm15[0,1,2],ymm8[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm5 = mem[0,1,2,3,4,5,6],ymm5[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm2[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[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm12[0,1,2,3,4,5,6],ymm6[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm1[5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm7 = ymm7[0,1,2,3],mem[4,5],ymm7[6,7]
+; 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[3,4,5,6,7]
-; 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:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm1[5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[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[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm9[0,1,2,3,4,5,6],ymm4[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm10[5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0],mem[1],xmm5[2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = mem[0,1,2,3,4,5,6],ymm12[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
-; 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, 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, 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, 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, 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, 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:    vblendps {{.*#+}} ymm3 = ymm7[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm11[5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = mem[0,1,2,3],ymm4[4,5],mem[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[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm13[0,1,2,3,4,5,6],ymm9[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, (%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 32(%rsi)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 64(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, (%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 96(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 32(%rdx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, (%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 32(%rcx)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%r8)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%r8)
@@ -2033,11 +2027,11 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%r8)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%r8)
-; AVX1-ONLY-NEXT:    vmovaps %ymm4, (%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 64(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm2, 96(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%r9)
-; AVX1-ONLY-NEXT:    addq $984, %rsp # imm = 0x3D8
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 32(%r9)
+; AVX1-ONLY-NEXT:    addq $920, %rsp # imm = 0x398
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
@@ -3571,17 +3565,18 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; 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:    subq $2472, %rsp # imm = 0x9A8
+; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 576(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 512(%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:    vmovaps 480(%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 224(%rdi), %ymm6
 ; 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
@@ -3589,359 +3584,320 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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:    vblendps {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%rdi), %ymm1, %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 256(%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:    vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm6[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm7[2,3],ymm8[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
+; AVX1-ONLY-NEXT:    vmovaps 576(%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:    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:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm9[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 608(%rdi), %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; 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:    vblendps {{.*#+}} ymm2 = ymm0[0,1],ymm4[2,3],ymm0[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2],xmm3[3]
+; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %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:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 928(%rdi), %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; 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 1152(%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:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2],xmm3[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:    vmovaps 1216(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm13[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1248(%rdi), %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; 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:    vmovaps 32(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],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:    vmovaps (%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2],xmm3[3]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm11
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm11[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%rdi), %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; 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:    vmovaps 352(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],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:    vmovaps 320(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2],xmm3[3]
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm14[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%rdi), %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; 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:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2],xmm3[3]
+; AVX1-ONLY-NEXT:    vmovaps 704(%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 736(%rdi), %ymm15
+; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm0[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 768(%rdi), %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-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 {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm2[0],xmm3[1],xmm2[2],xmm3[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:    vmovaps 1056(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %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:    vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm1[4],ymm4[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 1088(%rdi), %ymm3, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-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:    vperm2f128 {{.*#+}} ymm3 = ymm5[2,3,0,1]
 ; 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:    vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm6[1,3],ymm3[6,5],ymm6[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 160(%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 192(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 304(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm0[1,3],ymm3[6,5],ymm0[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 480(%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:    vmovaps 512(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 624(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm7[2,3,0,1]
 ; 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:    vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm8[1,3],ymm3[6,5],ymm8[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm7[2,3],ymm8[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 800(%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:    vmovaps 832(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 944(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[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[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm12[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm13[1,3],ymm3[6,5],ymm13[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm13[0,1],ymm12[2,3],ymm13[4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT:    vmovaps %ymm13, %ymm6
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 1120(%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, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 1264(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
 ; 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 %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm10[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm11[1,3],ymm3[6,5],ymm11[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 144(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[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[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm9[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm14[1,3],ymm3[6,5],ymm14[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm14[0,1],ymm9[2,3],ymm14[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-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:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 464(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
 ; 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 %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm15[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm13[1,3],ymm3[6,5],ymm13[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm13[0,1],ymm15[2,3],ymm13[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
+; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-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:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 784(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[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[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 928(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm9[1,3],ymm3[6,5],ymm9[5,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm9[0,1],ymm2[2,3],ymm9[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[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), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 1104(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
 ; 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 {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[2,0],ymm0[7,4],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,1],ymm0[6,4],ymm1[6,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm1[1,0],ymm10[0,0],ymm1[5,4],ymm10[4,4]
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[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), %ymm1 # 32-byte Reload
 ; 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[3,4,5,6,7]
-; 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:    vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[2,0],ymm0[7,4],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,1],ymm0[6,4],ymm1[6,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[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[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm2
 ; 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:    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), %ymm1 # 32-byte Reload
 ; 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:    vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[2,0],ymm0[7,4],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,1],ymm0[6,4],ymm1[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]
@@ -3949,6 +3905,43 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 928(%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 %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 {{.*#+}} ymm0 = ymm0[3,0],ymm6[2,0],ymm0[7,4],ymm6[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm6[2,1],ymm0[6,4],ymm6[6,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[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[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,0],ymm7[0,0],ymm1[5,4],ymm7[4,4]
+; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; 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],ymm9[2,0],ymm0[7,4],ymm9[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm9[2,1],ymm0[6,4],ymm9[6,5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[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[3,4,5,6,7]
 ; 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]
@@ -3958,8 +3951,9 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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:    vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm11[2,0],ymm0[7,4],ymm11[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm11[2,1],ymm0[6,4],ymm11[6,5]
+; AVX1-ONLY-NEXT:    vmovaps %ymm11, %ymm4
 ; 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]
@@ -3967,15 +3961,15 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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[3,4,5,6,7]
-; 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:    vmovaps 128(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm3[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[1,0],ymm9[0,0],ymm3[5,4],ymm9[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:    vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm14[2,0],ymm0[7,4],ymm14[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm14[2,1],ymm1[6,4],ymm14[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]
@@ -3990,8 +3984,8 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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:    vshufps {{.*#+}} ymm2 = ymm0[3,0],ymm13[2,0],ymm0[7,4],ymm13[6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm13[2,1],ymm2[6,4],ymm13[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]
@@ -4000,24 +3994,24 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} xmm5 = xmm5[1,0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm2[3,4,5,6,7]
 ; 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:    vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm8[1,0],ymm1[0,0],ymm8[5,4],ymm1[4,4]
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-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:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm11[0,0],ymm1[3,0],ymm11[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[3,4,5,6,7]
 ; 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:    vshufps {{.*#+}} ymm15 = ymm1[2,0],ymm10[1,0],ymm1[6,4],ymm10[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
@@ -4025,9 +4019,9 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm5[3,0],ymm1[4,4],ymm5[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm5[2,2],ymm15[6,4],ymm5[6,6]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -4039,14 +4033,15 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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), %ymm10 # 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 = ymm10[0,0],ymm1[3,0],ymm10[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[3,4,5,6,7]
 ; 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:    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
@@ -4059,18 +4054,17 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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[3,4,5,6,7]
-; 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:    vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[2,0],ymm7[1,0],ymm1[6,4],ymm7[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), %ymm10 # 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 = ymm10[0,0],ymm1[3,0],ymm10[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[3,4,5,6,7]
@@ -4088,7 +4082,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm8[2,0],ymm3[1,0],ymm8[6,4],ymm3[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm8[2,0],ymm2[1,0],ymm8[6,4],ymm2[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
@@ -4096,8 +4090,8 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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 {{.*#+}} ymm15 = ymm2[0,0],ymm14[3,0],ymm2[4,4],ymm14[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm14[2,2],ymm15[6,4],ymm14[6,6]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm6[2,0],ymm12[1,0],ymm6[6,4],ymm12[5,4]
@@ -4108,14 +4102,14 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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 {{.*#+}} ymm15 = ymm1[0,0],ymm4[3,0],ymm1[4,4],ymm4[7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm4[2,2],ymm15[6,4],ymm4[6,6]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[2,0],ymm10[1,0],ymm4[6,4],ymm10[5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm3[2,0],ymm9[1,0],ymm3[6,4],ymm9[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:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],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
@@ -4123,11 +4117,11 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3,4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5,6],ymm3[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:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %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[0,1,2,3,4],ymm9[5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm11[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]
@@ -4135,10 +4129,10 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0],mem[1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; 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 $128, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = ymm1[0,1,2,3,4,5,6],mem[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:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm4 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],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
@@ -4178,21 +4172,21 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX1-ONLY-NEXT:    vblendps $32, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm5 = ymm5[0,1,2,3,4],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:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[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[3,4,5,6,7]
 ; 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 $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = mem[0,1,2,3,4,5,6],ymm6[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[0,1,2,3,4],ymm14[5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm10[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:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = ymm7[0,1,2,3],mem[4,5],ymm7[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[3,4,5,6,7]
@@ -4211,7 +4205,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0],mem[1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3,4,5,6,7]
 ; 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:    vblendps $128, (%rsp), %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
@@ -4285,7 +4279,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; 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:    addq $2472, %rsp # imm = 0x9A8
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;

diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll
index 440c7b7..77cd728 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll
@@ -401,46 +401,46 @@ define void @store_i32_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm4
 ; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm6
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm7
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm5 = zero,xmm7[1],xmm6[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm8, %ymm5
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm9
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm10 = xmm9[0],xmm8[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,2,0]
-; AVX1-ONLY-NEXT:    vbroadcastss 4(%rdx), %xmm11
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm10, %ymm10
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6],ymm10[7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r8), %ymm1, %ymm10
+; AVX1-ONLY-NEXT:    vmovaps 16(%rsi), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm5 = zero,xmm8[1],xmm6[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm10 = xmm8[0],xmm6[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm10, %ymm5
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm11
+; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm12
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm12[0],xmm11[0]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,0]
+; AVX1-ONLY-NEXT:    vbroadcastss 4(%rdx), %xmm14
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm13, %ymm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm13[2,3],ymm5[4,5,6],ymm13[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r8), %ymm10, %ymm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0],ymm5[1,2,3],ymm10[4],ymm5[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rsi), %xmm10
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm11
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm12 = zero,zero,xmm11[2],xmm10[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm11[1,1],xmm10[1,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm10, %ymm10
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm4[1,1],ymm2[1,1],ymm4[5,5],ymm2[5,5]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm11[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3,4],ymm10[5,6,7]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0],ymm10[1,2,3,4],ymm11[5],ymm10[6,7]
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
-; AVX1-ONLY-NEXT:    vbroadcastss 4(%rcx), %xmm9
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm9, %ymm8
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm6 = zero,zero,xmm7[2],xmm6[2]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm8[0,1],ymm6[2,3],ymm8[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm7 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3,4,5],ymm7[6],ymm6[7]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm7
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm8
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm8[3,3],xmm7[3,3]
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm7 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm7, %ymm7
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm3[3,3],ymm1[3,3],ymm3[7,7],ymm1[7,7]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm8[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3,4],ymm7[5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm11[2],ymm7[3,4,5,6],ymm11[7]
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm10 = zero,zero,xmm9[2],xmm7[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm9[1,1],xmm7[1,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm7, %ymm7
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm4[1,1],ymm2[1,1],ymm4[5,5],ymm2[5,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm9[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4],ymm7[5,6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0],ymm7[1,2,3,4],ymm9[5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm11[2],xmm12[2],xmm11[3],xmm12[3]
+; AVX1-ONLY-NEXT:    vbroadcastss 4(%rcx), %xmm11
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm11, %ymm10
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm6 = zero,zero,xmm8[2],xmm6[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm6, %ymm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm6[2,3],ymm10[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm8 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm8[1],ymm6[2,3,4,5],ymm8[6],ymm6[7]
+; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm10
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm10[3,3],xmm8[3,3]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm10[2],xmm8[2],xmm10[3],xmm8[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm8, %ymm8
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm3[3,3],ymm1[3,3],ymm3[7,7],ymm1[7,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm10[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm10[3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4,5,6],ymm9[7]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4],ymm3[5,6,7]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[0,1,3,0,4,5,7,4]
@@ -449,9 +449,9 @@ define void @store_i32_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm7, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm6, 32(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm10, 96(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 96(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm5, (%r9)
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
@@ -472,9 +472,9 @@ define void @store_i32_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm9
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0,1],ymm5[2,3],ymm10[4,5,6],ymm5[7]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, (%r8), %ymm1, %ymm10
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm11 = ymm10[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm11[0,1],ymm5[2,3],ymm11[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, (%r8), %ymm10, %ymm10
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0],ymm5[1,2,3],ymm10[4],ymm5[5,6,7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,2,2]
@@ -542,10 +542,10 @@ define void @store_i32_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm8, %ymm8
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm7 = <0,1,u,u,3,2,3,u>
-; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm7, %ymm6
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm8[2,3],ymm6[4,5,6],ymm8[7]
-; AVX2-FAST-NEXT:    vinsertf128 $1, (%r8), %ymm1, %ymm7
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0],ymm6[1,2,3],ymm7[4],ymm6[5,6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm7, %ymm7
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm8[2,3],ymm7[4,5,6],ymm8[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, (%r8), %ymm6, %ymm6
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4],ymm7[5,6,7]
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm4[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm2[1],ymm7[2,3,4],ymm2[5],ymm7[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,2]
@@ -595,9 +595,9 @@ define void @store_i32_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm9
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0,1],ymm5[2,3],ymm10[4,5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, (%r8), %ymm1, %ymm10
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm11 = ymm10[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm11[0,1],ymm5[2,3],ymm11[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, (%r8), %ymm10, %ymm10
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0],ymm5[1,2,3],ymm10[4],ymm5[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,2,2]
@@ -861,257 +861,252 @@ define void @store_i32_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride5_vf16:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    pushq %rax
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %ymm4
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm6
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm3 = zero,xmm6[1],xmm8[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm8[0],xmm6[1],xmm8[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm5, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm4 = zero,xmm10[1],xmm6[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm5 = xmm10[0],xmm6[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm5, %ymm4
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm11
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm12
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm12[0],xmm11[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,2,0]
-; AVX1-ONLY-NEXT:    vbroadcastss 4(%rdx), %xmm7
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm5, %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3],ymm3[4,5,6],ymm5[7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r8), %ymm1, %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0],ymm3[1,2,3],ymm5[4],ymm3[5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm10
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm9
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm5 = zero,xmm9[1],xmm10[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm7, %ymm5
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm12[0],xmm11[0]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,2,0]
+; AVX1-ONLY-NEXT:    vbroadcastss 4(%rdx), %xmm13
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm9, %ymm9
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm9[2,3],ymm4[4,5,6],ymm9[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r8), %ymm5, %ymm5
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2,3],ymm5[4],ymm4[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm5 = zero,xmm8[1],xmm7[1],zero
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm13
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm14
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm14[0],xmm13[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[0,1,2,0]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm14[0],xmm13[0]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 36(%rdx), %xmm15
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm7, %ymm15
-; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %ymm7
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm15[2,3],ymm5[4,5,6],ymm15[7]
-; AVX1-ONLY-NEXT:    vmovaps %ymm2, %ymm1
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r8), %ymm2, %ymm15
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm15[0],ymm5[1,2,3],ymm15[4],ymm5[5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm9, %ymm9
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm15 = xmm8[0],xmm7[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm15, %ymm5
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm9[2,3],ymm5[4,5,6],ymm9[7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %ymm9
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r8), %ymm15, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm15[0],ymm5[1,2,3],ymm15[4],ymm5[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm11 = xmm11[2],xmm12[2],xmm11[3],xmm12[3]
 ; AVX1-ONLY-NEXT:    vbroadcastss 4(%rcx), %xmm12
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm12, %ymm11
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm8 = zero,zero,xmm6[2],xmm8[2]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm8, %ymm6
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm6 = zero,zero,xmm10[2],xmm6[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm11[0,1],ymm6[2,3],ymm11[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm8 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0],ymm8[1],ymm6[2,3,4,5],ymm8[6],ymm6[7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 48(%rsi), %xmm8
+; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm10 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0],ymm10[1],ymm6[2,3,4,5],ymm10[6],ymm6[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 48(%rsi), %xmm10
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm11
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm12 = zero,zero,xmm11[2],xmm8[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm11[1,1],xmm8[1,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm8, %ymm8
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm7[1,1],ymm0[1,1],ymm7[5,5],ymm0[5,5]
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm12 = zero,zero,xmm11[2],xmm10[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm11[1,1],xmm10[1,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm10, %ymm10
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm9[1,1],ymm2[1,1],ymm9[5,5],ymm2[5,5]
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, %ymm6
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm11[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm11[3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3,4],ymm10[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm11
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rdx), %xmm12
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm12[3,3],xmm11[3,3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm11 = xmm12[2],xmm11[2],xmm12[3],xmm11[3]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm12
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm11, %ymm11
-; AVX1-ONLY-NEXT:    vmovaps %ymm4, %ymm2
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[3,3],ymm1[3,3],ymm4[7,7],ymm1[7,7]
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, %ymm5
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm3[3,3],ymm1[3,3],ymm3[7,7],ymm1[7,7]
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, %ymm2
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm15[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm15[3,4],ymm11[5,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm12[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm15[0],ymm8[1,2,3,4],ymm15[5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0],ymm10[1,2,3,4],ymm15[5],ymm10[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2],ymm11[3,4,5,6],ymm15[7]
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm15
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm13 = xmm13[2],xmm14[2],xmm13[3],xmm14[3]
 ; AVX1-ONLY-NEXT:    vbroadcastss 36(%rcx), %xmm14
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm14, %ymm13
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm10 = zero,zero,xmm9[2],xmm10[2]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm10, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm13[0,1],ymm9[2,3],ymm13[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm10 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2,3,4,5],ymm10[6],ymm9[7]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm10
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm7 = zero,zero,xmm8[2],xmm7[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm7, %ymm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm13[0,1],ymm7[2,3],ymm13[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm8 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3,4,5],ymm8[6],ymm7[7]
+; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm8
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm13
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm13[3,3],xmm10[3,3]
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm13[2],xmm10[2],xmm13[3],xmm10[3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm10, %ymm10
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm15[3,3],ymm4[3,3],ymm15[7,7],ymm4[7,7]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm13[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm13[3,4],ymm10[5,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm13[3,3],xmm8[3,3]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm13[2],xmm8[2],xmm13[3],xmm8[3]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rsi), %xmm13
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm8, %ymm8
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, %ymm5
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm15[3,3],ymm0[3,3],ymm15[7,7],ymm0[7,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm14[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm14[3,4],ymm8[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm14
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm0 = zero,zero,xmm14[2],xmm13[2]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm14[1,1],xmm13[1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm13, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm13
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm14
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm14[1,1],ymm13[1,1],ymm14[5,5],ymm13[5,5]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2],ymm3[3,4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm3
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm0[2],ymm10[3,4,5,6],ymm0[7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm14[1,1],ymm13[1,1],ymm14[5,5],ymm13[5,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2],ymm4[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm4
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm0[2],ymm8[3,4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm7[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm7 = ymm6[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm2[2],ymm7[3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm9[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm6[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4,5,6],ymm3[7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm12[3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm15[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm14[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm7 = ymm13[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1],ymm4[2],ymm7[3,4,5,6],ymm4[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1,2,3],ymm2[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm14[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm9 = ymm13[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm9[0,1],ymm3[2],ymm9[3,4,5,6],ymm3[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm2, 64(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm10, 128(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm9, 192(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 192(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm11, 288(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm8, 256(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm10, 256(%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, 160(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%r9)
-; AVX1-ONLY-NEXT:    popq %rax
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: store_i32_stride5_vf16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %ymm4
-; AVX2-SLOW-NEXT:    vmovaps (%rdi), %ymm0
-; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %ymm6
+; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %ymm5
 ; AVX2-SLOW-NEXT:    vmovaps (%rdx), %ymm1
-; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %ymm5
-; AVX2-SLOW-NEXT:    vmovaps (%rcx), %ymm3
-; AVX2-SLOW-NEXT:    vmovaps (%r8), %ymm2
-; AVX2-SLOW-NEXT:    vmovaps 32(%r8), %ymm7
+; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %ymm4
+; AVX2-SLOW-NEXT:    vmovaps (%rcx), %ymm2
+; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %ymm7
+; AVX2-SLOW-NEXT:    vmovaps (%r8), %ymm0
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 32(%r8), %ymm6
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm9
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %xmm11
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm10
-; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %xmm13
+; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %xmm12
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,0,2,1]
-; AVX2-SLOW-NEXT:    vmovaps (%rdx), %xmm12
+; AVX2-SLOW-NEXT:    vmovaps (%rdx), %xmm13
 ; AVX2-SLOW-NEXT:    vmovaps (%rcx), %xmm14
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm15 = xmm14[1,1,2,2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm12[2],xmm15[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm13[2],xmm15[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm15[0,1],ymm8[2,3],ymm15[4,5],ymm8[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm2[0,1,2,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm15[1],ymm8[2,3,4,5],ymm15[6],ymm8[7]
-; AVX2-SLOW-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm0[0,1,2,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0],ymm15[1],ymm8[2,3,4,5],ymm15[6],ymm8[7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %xmm15
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm12[0],xmm14[0],xmm12[1],xmm14[1]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %xmm14
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm12 = xmm12[0,1,2,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[0,0,2,1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm12[2,3],ymm9[4,5,6],ymm12[7]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, (%r8), %ymm0, %ymm10
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0],ymm9[1,2,3],ymm10[4],ymm9[5,6,7]
-; AVX2-SLOW-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm13[2],xmm11[2],xmm13[3],xmm11[3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm9[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm13[2,3],ymm10[4,5,6],ymm13[7]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, (%r8), %ymm9, %ymm9
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0],ymm10[1,2,3],ymm9[4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm12[2],xmm11[2],xmm12[3],xmm11[3]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm12 = xmm14[1,1,2,2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm12 = xmm12[0,1],xmm15[2],xmm12[3]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[0,1,1,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3],ymm12[4,5],ymm10[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm7[0,1,2,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0],ymm12[1],ymm10[2,3,4,5],ymm12[6],ymm10[7]
-; AVX2-SLOW-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %ymm12
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[0,1,2,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[0,0,2,1]
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm13[0],xmm11[0],xmm13[1],xmm11[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm14[1,1,2,2]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm15[2],xmm13[3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,1,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm13[0,1],ymm10[2,3],ymm13[4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm6[0,1,2,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm13[1],ymm10[2,3,4,5],ymm13[6],ymm10[7]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,0,2,1]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm14[2,3],ymm11[4,5,6],ymm14[7]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, 32(%r8), %ymm4, %ymm13
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm13[0],ymm11[1,2,3],ymm13[4],ymm11[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm12[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm5[1],ymm13[2,3,4],ymm5[5],ymm13[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,2]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm6[1,1,2,2,5,5,6,6]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0],ymm4[1,2],ymm14[3,4],ymm4[5,6],ymm14[7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,2,2,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3],ymm12[4,5,6],ymm13[7]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, 32(%r8), %ymm11, %ymm11
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm12[1,2,3],ymm11[4],ymm12[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm12 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm4[1],ymm12[2,3,4],ymm4[5],ymm12[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[2,2,2,2]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm3[1,2],ymm13[3,4],ymm3[5,6],ymm13[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4],ymm13[5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastsd 48(%r8), %ymm13
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0],ymm12[1,2,3,4],ymm13[5],ymm12[6,7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm13 = ymm5[2],ymm3[2],ymm5[3],ymm3[3],ymm5[6],ymm3[6],ymm5[7],ymm3[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[3,3,3,3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm7[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1,2],ymm15[3,4],ymm14[5,6],ymm15[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4],ymm14[5,6,7]
-; AVX2-SLOW-NEXT:    vbroadcastsd 48(%r8), %ymm14
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0],ymm13[1,2,3,4],ymm14[5],ymm13[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm12[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastsd 56(%r8), %ymm14
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2],ymm13[3,4,5,6],ymm14[7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm1[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1,2],ymm15[3,4],ymm14[5,6],ymm15[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm6[2],ymm4[2],ymm6[3],ymm4[3],ymm6[6],ymm4[6],ymm6[7],ymm4[7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
+; AVX2-SLOW-NEXT:    vmovaps (%rdi), %ymm15
+; AVX2-SLOW-NEXT:    vmovaps (%rsi), %ymm0
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm8 = ymm0[2],ymm15[2],ymm0[3],ymm15[3],ymm0[6],ymm15[6],ymm0[7],ymm15[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,1,3,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4],ymm14[5,6,7]
-; AVX2-SLOW-NEXT:    vbroadcastsd 56(%r8), %ymm15
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3,4,5,6],ymm15[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm1[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm15[1,2],ymm8[3,4],ymm15[5,6],ymm8[7]
-; AVX2-SLOW-NEXT:    vmovaps (%rsi), %ymm15
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm9 = ymm15[2],ymm0[2],ymm15[3],ymm0[3],ymm15[6],ymm0[6],ymm15[7],ymm0[7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,1,3,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4],ymm8[5,6,7]
-; AVX2-SLOW-NEXT:    vbroadcastsd 24(%r8), %ymm9
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4,5,6],ymm9[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm3[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm1[1],ymm9[2,3,4],ymm1[5],ymm9[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,2,2,2]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm15[1,1,2,2,5,5,6,6]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm0[1,2],ymm10[3,4],ymm0[5,6],ymm10[7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4],ymm10[5,6,7]
-; AVX2-SLOW-NEXT:    vbroadcastsd 16(%r8), %ymm10
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0],ymm9[1,2,3,4],ymm10[5],ymm9[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm12[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,3,0,1,6,7,4,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2],ymm5[3,4,5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm7[3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm15[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm14[0,1,2],ymm8[3,4],ymm14[5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastsd 24(%r8), %ymm14
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm14[2],ymm8[3,4,5,6],ymm14[7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0],ymm1[1],ymm14[2,3,4],ymm1[5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,2,2,2]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm15[1,2],ymm9[3,4],ymm15[5,6],ymm9[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,2,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm14[3,4],ymm9[5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastsd 16(%r8), %ymm14
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm14[0],ymm9[1,2,3,4],ymm14[5],ymm9[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm7[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,3,0,4,5,7,4]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps %ymm0, 64(%r9)
-; AVX2-SLOW-NEXT:    vmovaps %ymm4, 224(%r9)
+; AVX2-SLOW-NEXT:    vmovaps %ymm3, 224(%r9)
 ; AVX2-SLOW-NEXT:    vmovaps %ymm9, 96(%r9)
 ; AVX2-SLOW-NEXT:    vmovaps %ymm8, 128(%r9)
-; AVX2-SLOW-NEXT:    vmovaps %ymm14, 288(%r9)
-; AVX2-SLOW-NEXT:    vmovaps %ymm13, 256(%r9)
+; AVX2-SLOW-NEXT:    vmovaps %ymm13, 288(%r9)
+; AVX2-SLOW-NEXT:    vmovaps %ymm12, 256(%r9)
 ; AVX2-SLOW-NEXT:    vmovaps %ymm11, 160(%r9)
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vmovaps %ymm0, 192(%r9)
+; AVX2-SLOW-NEXT:    vmovaps %ymm10, 192(%r9)
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps %ymm0, (%r9)
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -1121,114 +1116,114 @@ define void @store_i32_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX2-FAST-LABEL: store_i32_stride5_vf16:
 ; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %ymm2
-; AVX2-FAST-NEXT:    vmovaps (%rdi), %ymm1
-; AVX2-FAST-NEXT:    vmovaps (%r8), %ymm3
-; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 32(%r8), %ymm4
-; AVX2-FAST-NEXT:    vmovaps (%rsi), %xmm0
+; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %ymm2
+; AVX2-FAST-NEXT:    vmovaps (%r8), %ymm0
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 32(%r8), %ymm3
+; AVX2-FAST-NEXT:    vmovaps (%rsi), %xmm6
 ; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %xmm7
-; AVX2-FAST-NEXT:    vmovaps (%rdi), %xmm6
+; AVX2-FAST-NEXT:    vmovaps (%rdi), %xmm10
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %xmm9
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm6[2],xmm0[2],xmm6[3],xmm0[3]
-; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm12 = <0,1,0,1,u,u,2,2>
-; AVX2-FAST-NEXT:    vpermps %ymm5, %ymm12, %ymm5
-; AVX2-FAST-NEXT:    vmovaps (%rdx), %xmm10
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm10[2],xmm6[2],xmm10[3],xmm6[3]
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm11 = <0,1,0,1,u,u,2,2>
+; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm11, %ymm4
+; AVX2-FAST-NEXT:    vmovaps (%rdx), %xmm12
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %xmm13
-; AVX2-FAST-NEXT:    vmovaps (%rcx), %xmm11
+; AVX2-FAST-NEXT:    vmovaps (%rcx), %xmm14
 ; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %xmm15
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm8 = xmm11[1,1,2,2]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm10[2],xmm8[3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3],ymm8[4,5],ymm5[6,7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm3[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0],ymm8[1],ymm5[2,3,4,5],ymm8[6],ymm5[7]
-; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm5 = xmm14[1,1,2,2]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm12[2],xmm5[3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm0[0,1,2,1]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0],ymm5[1],ymm4[2,3,4,5],ymm5[6],ymm4[7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm7[2],xmm9[3],xmm7[3]
-; AVX2-FAST-NEXT:    vpermps %ymm5, %ymm12, %ymm5
+; AVX2-FAST-NEXT:    vpermps %ymm5, %ymm11, %ymm5
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm8 = xmm15[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm13[2],xmm8[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3],ymm8[4,5],ymm5[6,7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm4[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0],ymm8[1],ymm5[2,3,4,5],ymm8[6],ymm5[7]
-; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm3[0,1,2,1]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0],ymm8[1],ymm5[2,3,4,5],ymm8[6],ymm5[7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %ymm8
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm10[0],xmm11[0],xmm10[1],xmm11[1]
-; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %ymm11
-; AVX2-FAST-NEXT:    vpermps %ymm10, %ymm12, %ymm10
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
-; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm3 = [0,1,3,2,3,2,3,2]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm3, %ymm0
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm10[2,3],ymm0[4,5,6],ymm10[7]
-; AVX2-FAST-NEXT:    vinsertf128 $1, (%r8), %ymm1, %ymm6
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm0[1,2,3],ymm6[4],ymm0[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1],ymm0[2,3,4],ymm8[5],ymm0[6,7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm0[2,2,2,2]
-; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm14 = ymm0[1,1,2,2,5,5,6,6]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0],ymm2[1,2],ymm14[3,4],ymm2[5,6],ymm14[7]
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm12[0],xmm14[0],xmm12[1],xmm14[1]
+; AVX2-FAST-NEXT:    vpermps %ymm12, %ymm11, %ymm12
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm10[0],xmm6[0],xmm10[1],xmm6[1]
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm0 = [0,1,3,2,3,2,3,2]
+; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm0, %ymm10
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm12[2,3],ymm10[4,5,6],ymm12[7]
+; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %ymm12
+; AVX2-FAST-NEXT:    vinsertf128 $1, (%r8), %ymm6, %ymm6
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm10[1,2,3],ymm6[4],ymm10[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm8[1],ymm10[2,3,4],ymm8[5],ymm10[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,2]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm14 = ymm2[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0],ymm1[1,2],ymm14[3,4],ymm1[5,6],ymm14[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm14[0,1,2],ymm10[3,4],ymm14[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 48(%r8), %ymm14
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm14[0],ymm10[1,2,3,4],ymm14[5],ymm10[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm14 = ymm11[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm8[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm14[1,2],ymm5[3,4],ymm14[5,6],ymm5[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm14 = ymm0[2],ymm2[2],ymm0[3],ymm2[3],ymm0[6],ymm2[6],ymm0[7],ymm2[7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm14 = ymm12[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm8[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm14[1,2],ymm4[3,4],ymm14[5,6],ymm4[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm14 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,1,3,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm14[3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[2,1,3,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm14[3,4],ymm4[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 56(%r8), %ymm14
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm5[0,1],ymm14[2],ymm5[3,4,5,6],ymm14[7]
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm13[0],xmm15[0],xmm13[1],xmm15[1]
-; AVX2-FAST-NEXT:    vpermps %ymm5, %ymm12, %ymm5
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm4[0,1],ymm14[2],ymm4[3,4,5,6],ymm14[7]
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm13[0],xmm15[0],xmm13[1],xmm15[1]
+; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm11, %ymm4
+; AVX2-FAST-NEXT:    vmovaps (%rdx), %ymm11
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm9[0],xmm7[0],xmm9[1],xmm7[1]
-; AVX2-FAST-NEXT:    vmovaps (%rdx), %ymm9
-; AVX2-FAST-NEXT:    vpermps %ymm7, %ymm3, %ymm3
-; AVX2-FAST-NEXT:    vmovaps (%rcx), %ymm12
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3],ymm3[4,5,6],ymm5[7]
-; AVX2-FAST-NEXT:    vinsertf128 $1, 32(%r8), %ymm2, %ymm5
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm5[0],ymm3[1,2,3],ymm5[4],ymm3[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm12[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm9[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0],ymm3[1,2],ymm5[3,4],ymm3[5,6],ymm5[7]
-; AVX2-FAST-NEXT:    vmovaps (%rsi), %ymm5
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm13 = ymm5[2],ymm1[2],ymm5[3],ymm1[3],ymm5[6],ymm1[6],ymm5[7],ymm1[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[2,1,3,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 24(%r8), %ymm13
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm13[2],ymm3[3,4,5,6],ymm13[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm13 = ymm12[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm9[1],ymm13[2,3,4],ymm9[5],ymm13[6,7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,2]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm5[1,1,2,2,5,5,6,6]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm1[1,2],ymm15[3,4],ymm1[5,6],ymm15[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT:    vmovaps (%rcx), %ymm9
+; AVX2-FAST-NEXT:    vpermps %ymm7, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3],ymm0[4,5,6],ymm4[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, 32(%r8), %ymm7, %ymm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm4[0],ymm0[1,2,3],ymm4[4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm9[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm11[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3,4],ymm0[5,6],ymm4[7]
+; AVX2-FAST-NEXT:    vmovaps (%rdi), %ymm4
+; AVX2-FAST-NEXT:    vmovaps (%rsi), %ymm13
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm13[2],ymm4[2],ymm13[3],ymm4[3],ymm13[6],ymm4[6],ymm13[7],ymm4[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 24(%r8), %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm15[2],ymm0[3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm11[1],ymm15[2,3,4],ymm11[5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,2]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm13[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm4[1,2],ymm5[3,4],ymm4[5,6],ymm5[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm15[3,4],ymm5[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 16(%r8), %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0],ymm13[1,2,3,4],ymm15[5],ymm13[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4],ymm0[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm11[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0,1],ymm2[2],ymm8[3,4,5,6],ymm2[7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm5[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm15[0],ymm5[1,2,3,4],ymm15[5],ymm5[6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4],ymm2[5,6,7]
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm12[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm9[2,3,0,1,6,7,4,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2],ymm4[3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0,1],ymm2[2],ymm8[3,4,5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm1 = ymm1[0,1,2],mem[3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT:    vmovaps %ymm1, 64(%r9)
-; AVX2-FAST-NEXT:    vmovaps %ymm0, 224(%r9)
-; AVX2-FAST-NEXT:    vmovaps %ymm13, 96(%r9)
-; AVX2-FAST-NEXT:    vmovaps %ymm3, 128(%r9)
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm13[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm9[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm11[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4,5],ymm3[6,7]
+; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovaps %ymm2, 64(%r9)
+; AVX2-FAST-NEXT:    vmovaps %ymm1, 224(%r9)
+; AVX2-FAST-NEXT:    vmovaps %ymm5, 96(%r9)
+; AVX2-FAST-NEXT:    vmovaps %ymm0, 128(%r9)
 ; AVX2-FAST-NEXT:    vmovaps %ymm7, 160(%r9)
 ; AVX2-FAST-NEXT:    vmovaps %ymm14, 288(%r9)
 ; AVX2-FAST-NEXT:    vmovaps %ymm10, 256(%r9)
@@ -1242,122 +1237,122 @@ define void @store_i32_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX2-FAST-PERLANE-LABEL: store_i32_stride5_vf16:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %ymm3
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r8), %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r8), %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r8), %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r8), %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm9
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %xmm11
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm10
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %xmm13
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %xmm12
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %xmm13
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %xmm14
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm15 = xmm14[1,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm12[2],xmm15[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm13[2],xmm15[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm15[0,1],ymm8[2,3],ymm15[4,5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm2[0,1,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm15[1],ymm8[2,3,4,5],ymm15[6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm0[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0],ymm15[1],ymm8[2,3,4,5],ymm15[6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %xmm15
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm12[0],xmm14[0],xmm12[1],xmm14[1]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %xmm14
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm12 = xmm12[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm12[2,3],ymm9[4,5,6],ymm12[7]
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, (%r8), %ymm0, %ymm10
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0],ymm9[1,2,3],ymm10[4],ymm9[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm13[2],xmm11[2],xmm13[3],xmm11[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm9[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm13[2,3],ymm10[4,5,6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, (%r8), %ymm9, %ymm9
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0],ymm10[1,2,3],ymm9[4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm12[2],xmm11[2],xmm12[3],xmm11[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm12 = xmm14[1,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm12 = xmm12[0,1],xmm15[2],xmm12[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[0,1,1,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3],ymm12[4,5],ymm10[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm7[0,1,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0],ymm12[1],ymm10[2,3,4,5],ymm12[6],ymm10[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %ymm12
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm13[0],xmm11[0],xmm13[1],xmm11[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm14[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm15[2],xmm13[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm13[0,1],ymm10[2,3],ymm13[4,5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm6[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm13[1],ymm10[2,3,4,5],ymm13[6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm14[2,3],ymm11[4,5,6],ymm14[7]
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 32(%r8), %ymm4, %ymm13
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm13[0],ymm11[1,2,3],ymm13[4],ymm11[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm12[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm5[1],ymm13[2,3,4],ymm5[5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm6[1,1,2,2,5,5,6,6]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0],ymm4[1,2],ymm14[3,4],ymm4[5,6],ymm14[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3],ymm12[4,5,6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 32(%r8), %ymm11, %ymm11
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm12[1,2,3],ymm11[4],ymm12[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm12 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm4[1],ymm12[2,3,4],ymm4[5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm3[1,2],ymm13[3,4],ymm3[5,6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4],ymm13[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 48(%r8), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0],ymm12[1,2,3,4],ymm13[5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm13 = ymm5[2],ymm3[2],ymm5[3],ymm3[3],ymm5[6],ymm3[6],ymm5[7],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm7[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1,2],ymm15[3,4],ymm14[5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4],ymm14[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 48(%r8), %ymm14
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0],ymm13[1,2,3,4],ymm14[5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm12[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 56(%r8), %ymm14
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2],ymm13[3,4,5,6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm1[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1,2],ymm15[3,4],ymm14[5,6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm6[2],ymm4[2],ymm6[3],ymm4[3],ymm6[6],ymm4[6],ymm6[7],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm8 = ymm0[2],ymm15[2],ymm0[3],ymm15[3],ymm0[6],ymm15[6],ymm0[7],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,1,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4],ymm14[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 56(%r8), %ymm15
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3,4,5,6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm1[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm15[1,2],ymm8[3,4],ymm15[5,6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %ymm15
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm9 = ymm15[2],ymm0[2],ymm15[3],ymm0[3],ymm15[6],ymm0[6],ymm15[7],ymm0[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,1,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4],ymm8[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 24(%r8), %ymm9
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4,5,6],ymm9[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm3[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm1[1],ymm9[2,3,4],ymm1[5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm15[1,1,2,2,5,5,6,6]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm0[1,2],ymm10[3,4],ymm0[5,6],ymm10[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4],ymm10[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 16(%r8), %ymm10
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0],ymm9[1,2,3,4],ymm10[5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm12[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,3,0,1,6,7,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2],ymm5[3,4,5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm7[3],ymm4[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm15[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm14[0,1,2],ymm8[3,4],ymm14[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 24(%r8), %ymm14
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm14[2],ymm8[3,4,5,6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0],ymm1[1],ymm14[2,3,4],ymm1[5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm15[1,2],ymm9[3,4],ymm15[5,6],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm14[3,4],ymm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 16(%r8), %ymm14
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm14[0],ymm9[1,2,3,4],ymm14[5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm7[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,3,0,4,5,7,4]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 64(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm4, 224(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm3, 224(%r9)
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 96(%r9)
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, 128(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm14, 288(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm13, 256(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm13, 288(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm12, 256(%r9)
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm11, 160(%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:    vmovaps %ymm10, 192(%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:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -1915,277 +1910,268 @@ define void @store_i32_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride5_vf32:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $728, %rsp # imm = 0x2D8
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm6
-; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm4
-; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm9
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm8
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm0 = zero,xmm8[1],xmm9[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
+; AVX1-ONLY-NEXT:    subq $616, %rsp # imm = 0x268
+; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm0 = zero,xmm9[1],xmm5[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = xmm9[0],xmm5[0],zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm15
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm5
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm5[0],xmm15[0]
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm12
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm12[0],xmm10[0]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 4(%rdx), %xmm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5,6],ymm2[7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r8), %ymm4, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3],ymm2[4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r8), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm4
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm0 = zero,xmm2[1],xmm4[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm3, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm0 = zero,xmm11[1],xmm7[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = xmm11[0],xmm7[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm4
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,0]
-; AVX1-ONLY-NEXT:    vbroadcastss 36(%rdx), %xmm7
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3],ymm0[4,5,6],ymm3[7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r8), %ymm6, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3],ymm3[4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm3[0],xmm4[0]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; AVX1-ONLY-NEXT:    vbroadcastss 36(%rdx), %xmm6
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r8), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm6
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm7 = zero,xmm6[1],xmm3[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm10, %ymm11
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm7
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm12 = xmm7[0],xmm10[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm12[0,1,2,0]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm8
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm8[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 68(%rdx), %xmm13
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm12, %ymm12
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm12[2,3],ymm11[4,5,6],ymm12[7]
-; 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, 64(%r8), %ymm0, %ymm12
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0],ymm11[1,2,3],ymm12[4],ymm11[5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm6, %ymm6
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm13 = zero,xmm1[1],xmm2[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm14 = xmm1[0],xmm2[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm14, %ymm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm13[0,1],ymm6[2,3],ymm13[4,5,6],ymm6[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r8), %ymm14, %ymm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0],ymm6[1,2,3],ymm13[4],ymm6[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm13
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm12
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm11 = zero,xmm12[1],xmm13[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm12[0],xmm13[0],xmm12[1],xmm13[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm14, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm11
-; AVX1-ONLY-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm6[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,0]
+; AVX1-ONLY-NEXT:    vbroadcastss 100(%rdx), %xmm13
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm6, %ymm13
+; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm0[0],xmm11[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[0,1,2,0]
-; AVX1-ONLY-NEXT:    vbroadcastss 100(%rdx), %xmm11
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm14, %ymm11
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm11[2,3],ymm1[4,5,6],ymm11[7]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm11
-; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r8), %ymm11, %ymm11
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0],ymm0[1,2,3],ymm11[4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm15 = zero,xmm14[1],xmm0[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm0 = xmm14[0],xmm0[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1],ymm13[2,3],ymm15[4,5,6],ymm13[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r8), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1,2,3],ymm0[4],ymm13[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm15[2],xmm5[2],xmm15[3],xmm5[3]
-; AVX1-ONLY-NEXT:    vbroadcastss 4(%rcx), %xmm5
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm5, %ymm0
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm5 = zero,zero,xmm8[2],xmm9[2]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm5, %ymm5
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm10[2],xmm12[2],xmm10[3],xmm12[3]
+; AVX1-ONLY-NEXT:    vbroadcastss 4(%rcx), %xmm10
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm10, %ymm0
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm5 = zero,zero,xmm9[2],xmm5[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3],ymm0[4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm5 = mem[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3,4,5],ymm5[6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm5
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm8 = zero,zero,xmm5[2],xmm0[2]
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm9 = zero,zero,xmm5[2],xmm0[2]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm5[1,1],xmm0[1,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm13
+; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm5
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm8
-; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm8[1,1],ymm5[1,1],ymm8[5,5],ymm5[5,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,1],ymm13[1,1],ymm5[5,5],ymm13[5,5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm8
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm8[3,3],xmm5[3,3]
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm8[2],xmm5[2],xmm8[3],xmm5[3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm5, %ymm5
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm9 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm9
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[3,3],xmm5[3,3]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm5[2],xmm9[3],xmm5[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm5, %ymm5
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm9
+; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[3,3],ymm6[3,3],ymm9[7,7],ymm6[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm9[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm9[3,4],ymm5[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0],ymm0[1,2,3,4],ymm11[5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm6
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm6[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0],ymm0[1,2,3,4],ymm9[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm11[2],ymm5[3,4,5,6],ymm11[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm9[2],ymm5[3,4,5,6],ymm9[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, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vbroadcastss 36(%rcx), %xmm5
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm5, %ymm0
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm4 = zero,zero,xmm2[2],xmm4[2]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4,5],ymm2[6],ymm0[7]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX1-ONLY-NEXT:    vbroadcastss 36(%rcx), %xmm3
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm3, %ymm0
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm3 = zero,zero,xmm11[2],xmm7[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm3, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3],ymm0[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4,5],ymm3[6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rsi), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm4 = zero,zero,xmm2[2],xmm0[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[1,1],xmm0[1,1]
+; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm4 = zero,zero,xmm3[2],xmm0[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm0[1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,1],ymm1[1,1],ymm2[5,5],ymm1[5,5]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %ymm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[1,1],ymm3[1,1],ymm4[5,5],ymm3[5,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rdx), %xmm4
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm4[3,3],xmm2[3,3]
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm2, %ymm4
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm1[3,3],ymm8[3,3],ymm1[7,7],ymm8[7,7]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm11[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm4[0,1,2],ymm11[3,4],ymm4[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0],ymm0[1,2,3,4],ymm14[5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm4[3,3],xmm3[3,3]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %ymm12
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm12[3,3],ymm4[3,3],ymm12[7,7],ymm4[7,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm4
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0],ymm0[1,2,3,4],ymm4[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm14[2],ymm11[3,4,5,6],ymm14[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm4[2],ymm3[3,4,5,6],ymm4[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm10[2],xmm7[2],xmm10[3],xmm7[3]
-; AVX1-ONLY-NEXT:    vbroadcastss 68(%rcx), %xmm7
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm7, %ymm0
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm3 = zero,zero,xmm6[2],xmm3[2]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm3, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm8[2],xmm0[3],xmm8[3]
+; AVX1-ONLY-NEXT:    vbroadcastss 68(%rcx), %xmm3
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm3, %ymm0
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,zero,xmm1[2],xmm2[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm3 = zero,zero,xmm1[2],xmm0[2]
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,zero,xmm1[2],xmm0[2]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,1],xmm0[1,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm15
-; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %ymm7
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm7[1,1],ymm15[1,1],ymm7[5,5],ymm15[5,5]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2],ymm3[3,4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rcx), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm10[3,3],xmm0[3,3]
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm10[2],xmm0[2],xmm10[3],xmm0[3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm10
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %ymm4
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm4[3,3],ymm5[3,3],ymm4[7,7],ymm5[7,7]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm11[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm10[0,1,2],ymm11[3,4],ymm10[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm10
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm10[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0],ymm3[1,2,3,4],ymm14[5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm14[2],ymm11[3,4,5,6],ymm14[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm11
+; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %ymm9
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[1,1],ymm11[1,1],ymm9[5,5],ymm11[5,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 80(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm2
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[3,3],xmm1[3,3]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %ymm6
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[3,3],ymm7[3,3],ymm6[7,7],ymm7[7,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm3[0,1,2],ymm5[3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm8
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm8[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0],ymm4[1,2,3,4],ymm0[5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm0[2],ymm5[3,4,5,6],ymm0[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, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm11 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vbroadcastss 100(%rcx), %xmm14
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm14, %ymm11
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm13 = zero,zero,xmm12[2],xmm13[2]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm13, %ymm12
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm12[2,3],ymm11[4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm12 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0],ymm12[1],ymm11[2,3,4,5],ymm12[6],ymm11[7]
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vbroadcastss 100(%rcx), %xmm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm4, %ymm0
+; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm4 = zero,zero,xmm14[2],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm4, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3],ymm0[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm4 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4,5],ymm4[6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 112(%rsi), %xmm12
-; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm13
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm14 = zero,zero,xmm13[2],xmm12[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm13[1,1],xmm12[1,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm12, %ymm14
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %ymm12
-; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %ymm13
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm13[1,1],ymm12[1,1],ymm13[5,5],ymm12[5,5]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm14[0,1,2],ymm3[3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 112(%rcx), %xmm14
+; AVX1-ONLY-NEXT:    vmovaps 112(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm14 = zero,zero,xmm5[2],xmm0[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm5[1,1],xmm0[1,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm5
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %ymm14
+; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %ymm3
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm3[1,1],ymm14[1,1],ymm3[5,5],ymm14[5,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm15[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm15[3,4],ymm5[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 112(%rcx), %xmm15
 ; AVX1-ONLY-NEXT:    vmovaps 112(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3],xmm14[3,3]
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm14[2],xmm0[3],xmm14[3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3],xmm15[3,3]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %ymm2
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm2[3,3],ymm3[3,3],ymm2[7,7],ymm3[7,7]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm14[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2],ymm14[3,4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %ymm14
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm14[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm0[0],ymm9[1,2,3,4],ymm0[5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm1[0,1],ymm0[2],ymm1[3,4,5,6],ymm0[7]
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %ymm15
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm15[3,3],ymm2[3,3],ymm15[7,7],ymm2[7,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2],ymm4[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %ymm4
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm0[0],ymm5[1,2,3,4],ymm0[5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm1[0,1],ymm0[2],ymm1[3,4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = mem[3,0,2,3,7,4,6,7]
 ; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm11 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm0[2],ymm11[3,4,5,6],ymm0[7]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm13 = ymm13[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm0[2],ymm13[3,4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm1 = mem[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm8 = mem[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm11 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0,1],ymm8[2],ymm11[3,4,5,6],ymm8[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm8[1,2,3],ymm1[4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm12 = mem[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm13 = mem[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2],ymm13[3,4,5,6],ymm12[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm12[1,2,3],ymm1[4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0,1,2],mem[3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm7[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm7 = ymm15[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1],ymm5[2],ymm7[3,4,5,6],ymm5[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm10[3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm13[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm7 = ymm12[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1],ymm5[2],ymm7[3,4,5,6],ymm5[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1,2,3],ymm2[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm14[3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4],ymm6[5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm9[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm9 = ymm11[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm7[2],ymm9[3,4,5,6],ymm7[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm15[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4],ymm7[5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm7 = ymm14[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm7[0,1],ymm3[2],ymm7[3,4,5,6],ymm3[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm2, 544(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm4, 384(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm6, 384(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm6, 608(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm9, 576(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 608(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm10, 576(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 512(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 416(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 352(%r9)
@@ -2209,92 +2195,89 @@ define void @store_i32_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; 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, (%r9)
-; AVX1-ONLY-NEXT:    addq $728, %rsp # imm = 0x2D8
+; AVX1-ONLY-NEXT:    addq $616, %rsp # imm = 0x268
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: store_i32_stride5_vf32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    subq $552, %rsp # imm = 0x228
-; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %ymm12
-; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %ymm13
-; AVX2-SLOW-NEXT:    vmovaps (%rdi), %ymm15
-; AVX2-SLOW-NEXT:    vmovaps (%r8), %ymm14
+; AVX2-SLOW-NEXT:    vmovaps (%r8), %ymm15
+; AVX2-SLOW-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 32(%r8), %ymm14
 ; AVX2-SLOW-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 32(%r8), %ymm11
-; AVX2-SLOW-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 64(%r8), %ymm10
-; AVX2-SLOW-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps (%rcx), %xmm4
-; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %xmm2
+; AVX2-SLOW-NEXT:    vmovaps 64(%r8), %ymm13
+; AVX2-SLOW-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps (%rcx), %xmm7
+; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %xmm3
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rcx), %xmm0
-; AVX2-SLOW-NEXT:    vmovaps (%rdx), %xmm5
-; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %xmm3
+; AVX2-SLOW-NEXT:    vmovaps (%rdx), %xmm8
+; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %xmm4
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rdx), %xmm1
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
-; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm7
-; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm8
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0,1],ymm6[2,3],ymm9[4,5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, (%r8), %ymm15, %ymm9
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0],ymm6[1,2,3],ymm9[4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[0,0,2,1]
+; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm9
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %xmm6
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm7 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
-; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %xmm8
+; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %xmm2
+; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm10
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm12[0,1],ymm5[2,3],ymm12[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %xmm12
+; AVX2-SLOW-NEXT:    vinsertf128 $1, (%r8), %ymm11, %ymm11
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm11[0],ymm5[1,2,3],ymm11[4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %xmm5
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,2,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,0,2,1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[1,1,2,2]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2],xmm7[3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,1,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm15[0,1,2,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3,4,5],ymm8[6],ymm7[7]
+; AVX2-SLOW-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,1,2,2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,1,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3],ymm4[4,5],ymm7[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm14[0,1,2,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4,5],ymm5[6],ymm4[7]
-; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,0,2,1]
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm8[0],xmm6[0],xmm8[1],xmm6[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6],ymm4[7]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, 32(%r8), %ymm13, %ymm5
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2,3],ymm5[4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %xmm4
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %xmm6
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,2,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,1,2,2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2],xmm2[3]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,1,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm5[2,3],ymm2[4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm11[0,1,2,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3,4,5],ymm3[6],ymm2[7]
-; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, 64(%r8), %ymm12, %ymm3
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3],ymm3[4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm6[2],xmm4[2],xmm6[3],xmm4[3]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm12[0],xmm6[0],xmm12[1],xmm6[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm8[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6],ymm7[7]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, 32(%r8), %ymm8, %ymm8
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1,2,3],ymm8[4],ymm7[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm12[2],xmm6[2],xmm12[3],xmm6[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,1,2,2]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3],ymm3[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm14[0,1,2,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4,5],ymm4[6],ymm3[7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm4[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3],ymm6[4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, 64(%r8), %ymm4, %ymm4
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3],ymm4[4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm10[0,1,2,1]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm13[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rcx), %xmm0
@@ -2306,158 +2289,161 @@ define void @store_i32_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %ymm14
-; AVX2-SLOW-NEXT:    vinsertf128 $1, 96(%r8), %ymm14, %ymm5
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vmovaps (%rdi), %ymm6
+; AVX2-SLOW-NEXT:    vinsertf128 $1, 96(%r8), %ymm5, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps (%rdx), %ymm5
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX2-SLOW-NEXT:    vmovaps (%rcx), %ymm14
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
+; AVX2-SLOW-NEXT:    vmovaps 96(%r8), %ymm3
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vmovaps 96(%r8), %ymm1
-; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm3[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps (%rdx), %ymm11
-; AVX2-SLOW-NEXT:    vmovaps (%rcx), %ymm10
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm11[1],ymm0[2,3,4],ymm11[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3,4],ymm5[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovaps (%rsi), %ymm9
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[1,1,2,2,5,5,6,6]
-; AVX2-SLOW-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1,2],ymm1[3,4],ymm15[5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vmovaps (%rsi), %ymm13
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2],ymm1[3,4],ymm6[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 16(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm10[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm11[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm14[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm15[2],ymm9[3],ymm15[3],ymm9[6],ymm15[6],ymm9[7],ymm15[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm13[2],ymm6[2],ymm13[3],ymm6[3],ymm13[6],ymm6[6],ymm13[7],ymm6[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 24(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %ymm8
-; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %ymm7
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1],ymm0[2,3,4],ymm8[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %ymm12
+; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %ymm11
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %ymm6
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[1,1,2,2,5,5,6,6]
-; AVX2-SLOW-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1,2],ymm1[3,4],ymm13[5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %ymm10
+; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %ymm9
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1,2],ymm1[3,4],ymm10[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 48(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm13[2],ymm6[3],ymm13[3],ymm6[6],ymm13[6],ymm6[7],ymm13[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[6],ymm10[6],ymm9[7],ymm10[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 56(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 64(%rdx), %ymm5
-; AVX2-SLOW-NEXT:    vmovaps 64(%rcx), %ymm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm4[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3,4],ymm5[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 64(%rdx), %ymm8
+; AVX2-SLOW-NEXT:    vmovaps 64(%rcx), %ymm7
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1],ymm0[2,3,4],ymm8[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %ymm3
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[1,1,2,2,5,5,6,6]
-; AVX2-SLOW-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm12[1,2],ymm1[3,4],ymm12[5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %ymm6
+; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %ymm5
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2],ymm1[3,4],ymm6[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 80(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm4[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm12[2],ymm3[3],ymm12[3],ymm3[6],ymm12[6],ymm3[7],ymm12[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm6[2],ymm5[3],ymm6[3],ymm5[6],ymm6[6],ymm5[7],ymm6[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 88(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 96(%rdx), %ymm2
-; AVX2-SLOW-NEXT:    vmovaps 96(%rcx), %ymm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4],ymm2[5],ymm0[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovaps 96(%rsi), %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm0[1,1,2,2,5,5,6,6]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1,2],ymm13[3,4],ymm14[5,6],ymm13[7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4],ymm13[5,6,7]
-; AVX2-SLOW-NEXT:    vbroadcastsd 112(%r8), %ymm13
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm12[1,2,3,4],ymm13[5],ymm12[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm12 = ymm1[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm2[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0],ymm12[1,2],ymm15[3,4],ymm12[5,6],ymm15[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm0[2],ymm14[2],ymm0[3],ymm14[3],ymm0[6],ymm14[6],ymm0[7],ymm14[7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[2,1,3,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm15[3,4],ymm12[5,6,7]
-; AVX2-SLOW-NEXT:    vbroadcastsd 120(%r8), %ymm15
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm15[2],ymm12[3,4,5,6],ymm15[7]
+; AVX2-SLOW-NEXT:    vmovaps 96(%rdx), %ymm4
+; AVX2-SLOW-NEXT:    vmovaps 96(%rcx), %ymm3
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
+; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %ymm2
+; AVX2-SLOW-NEXT:    vmovaps 96(%rsi), %ymm1
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm2[1,2],ymm15[3,4],ymm2[5,6],ymm15[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2],ymm0[3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastsd 112(%r8), %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0],ymm0[1,2,3,4],ymm15[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0],ymm0[1,2],ymm15[3,4],ymm0[5,6],ymm15[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastsd 120(%r8), %ymm15
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0,1],ymm15[2],ymm0[3,4,5,6],ymm15[7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm13[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm14[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = mem[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1],ymm13[2],ymm14[3,4,5,6],ymm13[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1,2,3],ymm0[4,5],ymm13[6,7]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm9 = ymm9[0,1,2,3],mem[4],ymm9[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4],ymm9[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm11[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2],ymm11[3,4,5,6],ymm10[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4,5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm9 = ymm9[0,1,2],mem[3],ymm9[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm6 = ymm6[0,1,2,3],mem[4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2],ymm8[3,4,5,6],ymm7[7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4,5],ymm7[6,7]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm6 = ymm6[0,1,2],mem[3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm3 = ymm3[0,1,2,3],mem[4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,3,0,1,6,7,4,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3,4,5,6],ymm4[7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4,5],ymm4[6,7]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm14[4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT:    vmovaps %ymm0, 544(%r9)
-; AVX2-SLOW-NEXT:    vmovaps %ymm3, 384(%r9)
-; AVX2-SLOW-NEXT:    vmovaps %ymm6, 224(%r9)
-; AVX2-SLOW-NEXT:    vmovaps %ymm9, 64(%r9)
-; AVX2-SLOW-NEXT:    vmovaps %ymm12, 608(%r9)
-; AVX2-SLOW-NEXT:    vmovaps %ymm13, 576(%r9)
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm8[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2],ymm7[3,4,5,6],ymm6[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm5 = ymm5[0,1,2],mem[3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = ymm1[0,1,2],mem[3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps %ymm1, 544(%r9)
+; AVX2-SLOW-NEXT:    vmovaps %ymm5, 384(%r9)
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 224(%r9)
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 64(%r9)
+; AVX2-SLOW-NEXT:    vmovaps %ymm15, 608(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 576(%r9)
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps %ymm0, 448(%r9)
 ; AVX2-SLOW-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
@@ -2492,239 +2478,238 @@ define void @store_i32_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX2-FAST-LABEL: store_i32_stride5_vf32:
 ; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    subq $568, %rsp # imm = 0x238
+; AVX2-FAST-NEXT:    subq $584, %rsp # imm = 0x248
 ; AVX2-FAST-NEXT:    vmovaps (%r8), %ymm11
 ; AVX2-FAST-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 32(%r8), %ymm9
-; AVX2-FAST-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 32(%r8), %ymm10
+; AVX2-FAST-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps (%rsi), %xmm4
-; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %xmm10
+; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %xmm6
 ; AVX2-FAST-NEXT:    vmovaps 64(%rsi), %xmm15
 ; AVX2-FAST-NEXT:    vmovaps (%rdi), %xmm5
-; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %xmm2
-; AVX2-FAST-NEXT:    vmovaps 64(%rdi), %xmm13
-; AVX2-FAST-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %xmm9
+; AVX2-FAST-NEXT:    vmovaps 64(%rdi), %xmm12
+; AVX2-FAST-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm12 = <0,1,0,1,u,u,2,2>
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm12, %ymm0
-; AVX2-FAST-NEXT:    vmovaps (%rdx), %xmm6
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm14 = <0,1,0,1,u,u,2,2>
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
+; AVX2-FAST-NEXT:    vmovaps (%rdx), %xmm7
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps (%rcx), %xmm8
-; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %xmm3
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm7 = xmm8[1,1,2,2]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm6[2],xmm7[3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm0[2,3],ymm7[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm11[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3,4,5],ymm7[6],ymm0[7]
+; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %xmm2
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm8[1,1,2,2]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm7[2],xmm3[3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm11[0,1,2,1]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4,5],ymm3[6],ymm0[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm2[2],xmm10[2],xmm2[3],xmm10[3]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm12, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm7 = xmm3[1,1,2,2]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm1[2],xmm7[3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm0[2,3],ymm7[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm9[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3,4,5],ymm7[6],ymm0[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm9[2],xmm6[2],xmm9[3],xmm6[3]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm2[1,1,2,2]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2],xmm3[3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm10[0,1,2,1]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4,5],ymm3[6],ymm0[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm13[2],xmm15[2],xmm13[3],xmm15[3]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm12, %ymm9
-; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %xmm7
-; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %xmm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm11 = xmm0[1,1,2,2]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],xmm7[2],xmm11[3]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm12[2],xmm15[2],xmm12[3],xmm15[3]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm10
+; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %xmm0
+; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %xmm3
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm11 = xmm3[1,1,2,2]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],xmm0[2],xmm11[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1],ymm9[2,3],ymm11[4,5],ymm9[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 64(%r8), %ymm11
 ; AVX2-FAST-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1],ymm9[2,3,4,5],ymm11[6],ymm9[7]
-; AVX2-FAST-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm11[1],ymm10[2,3,4,5],ymm11[6],ymm10[7]
+; AVX2-FAST-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%rsi), %xmm11
 ; AVX2-FAST-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %xmm9
-; AVX2-FAST-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm9[2],xmm11[2],xmm9[3],xmm11[3]
-; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm12, %ymm9
-; AVX2-FAST-NEXT:    vmovaps 96(%rdx), %xmm13
-; AVX2-FAST-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %xmm14
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm11 = xmm14[1,1,2,2]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],xmm13[2],xmm11[3]
+; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %xmm10
+; AVX2-FAST-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm10[2],xmm11[2],xmm10[3],xmm11[3]
+; AVX2-FAST-NEXT:    vpermps %ymm10, %ymm14, %ymm10
+; AVX2-FAST-NEXT:    vmovaps 96(%rdx), %xmm12
+; AVX2-FAST-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %xmm11
+; AVX2-FAST-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[1,1,2,2]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],xmm12[2],xmm11[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1],ymm9[2,3],ymm11[4,5],ymm9[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 96(%r8), %ymm11
 ; AVX2-FAST-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1],ymm9[2,3,4,5],ymm11[6],ymm9[7]
-; AVX2-FAST-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm11[1],ymm10[2,3,4,5],ymm11[6],ymm10[7]
+; AVX2-FAST-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm12, %ymm5
-; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm6 = [0,1,3,2,3,2,3,2]
-; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm6, %ymm4
-; AVX2-FAST-NEXT:    vmovaps %ymm6, %ymm11
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3],ymm4[4,5,6],ymm5[7]
-; AVX2-FAST-NEXT:    vmovaps (%rdi), %ymm9
-; AVX2-FAST-NEXT:    vinsertf128 $1, (%r8), %ymm9, %ymm5
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2,3],ymm5[4],ymm4[5,6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm7, %ymm14, %ymm5
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm11 = [0,1,3,2,3,2,3,2]
+; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm11, %ymm7
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1],ymm5[2,3],ymm7[4,5,6],ymm5[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, (%r8), %ymm4, %ymm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps (%rdx), %ymm5
-; AVX2-FAST-NEXT:    vmovaps (%rcx), %ymm6
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm6[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vmovaps %ymm6, %ymm8
-; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps (%rcx), %ymm7
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vmovaps %ymm7, %ymm8
+; AVX2-FAST-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4],ymm5[5],ymm4[6,7]
-; AVX2-FAST-NEXT:    vmovaps %ymm5, %ymm6
+; AVX2-FAST-NEXT:    vmovaps %ymm5, %ymm7
 ; AVX2-FAST-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[2,2,2,2]
-; AVX2-FAST-NEXT:    vmovaps (%rsi), %ymm13
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm13[1,1,2,2,5,5,6,6]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm9[1,2],ymm5[3,4],ymm9[5,6],ymm5[7]
+; AVX2-FAST-NEXT:    vmovaps (%rdi), %ymm13
+; AVX2-FAST-NEXT:    vmovaps (%rsi), %ymm12
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm12[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm13[1,2],ymm5[3,4],ymm13[5,6],ymm5[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4],ymm5[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 16(%r8), %ymm5
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2,3,4],ymm5[5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm8[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm6[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm7[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2],ymm5[3,4],ymm4[5,6],ymm5[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm13[2],ymm9[2],ymm13[3],ymm9[3],ymm13[6],ymm9[6],ymm13[7],ymm9[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm12[2],ymm13[2],ymm12[3],ymm13[3],ymm12[6],ymm13[6],ymm12[7],ymm13[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[3,3,3,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[2,1,3,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4],ymm4[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 24(%r8), %ymm5
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4,5,6],ymm5[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
-; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm12, %ymm1
-; AVX2-FAST-NEXT:    vmovaps %ymm11, %ymm3
-; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm11, %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %ymm5
-; AVX2-FAST-NEXT:    vinsertf128 $1, 32(%r8), %ymm5, %ymm2
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm9[0],xmm6[0],xmm9[1],xmm6[1]
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm14, %ymm1
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm11, %ymm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3],ymm4[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, 32(%r8), %ymm2, %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %ymm11
-; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %ymm10
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %ymm10
+; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %ymm9
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1],ymm1[2,3,4],ymm10[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %ymm8
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm8[1,1,2,2,5,5,6,6]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1,2],ymm2[3,4],ymm5[5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %ymm8
+; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %ymm7
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm7[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm8[1,2],ymm2[3,4],ymm8[5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 48(%r8), %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3,4],ymm2[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm10[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm11[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm2 = ymm8[2],ymm5[2],ymm8[3],ymm5[3],ymm8[6],ymm5[6],ymm8[7],ymm5[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm2 = ymm7[2],ymm8[2],ymm7[3],ymm8[3],ymm7[6],ymm8[6],ymm7[7],ymm8[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[3,3,3,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,3,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 56(%r8), %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm12, %ymm0
-; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm3, %ymm1
-; AVX2-FAST-NEXT:    vmovaps %ymm3, %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6],ymm0[7]
-; AVX2-FAST-NEXT:    vmovaps 64(%rdi), %ymm3
-; AVX2-FAST-NEXT:    vinsertf128 $1, 64(%r8), %ymm3, %ymm1
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm11, %ymm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, 64(%r8), %ymm1, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %ymm7
-; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %ymm6
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3,4],ymm7[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %ymm6
+; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %ymm5
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm6[1],ymm0[2,3,4],ymm6[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-NEXT:    vmovaps 64(%rsi), %ymm4
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1,2],ymm1[3,4],ymm3[5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vmovaps 64(%rdi), %ymm4
+; AVX2-FAST-NEXT:    vmovaps 64(%rsi), %ymm3
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 80(%r8), %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[6],ymm3[6],ymm4[7],ymm3[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 88(%r8), %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
-; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm12, %ymm0
+; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-FAST-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
-; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm2, %ymm1
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6],ymm0[7]
-; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %ymm0
-; AVX2-FAST-NEXT:    vinsertf128 $1, 96(%r8), %ymm0, %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm11, %ymm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, 96(%r8), %ymm1, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%rdx), %ymm14
 ; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %ymm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm12 = ymm1[2,2,2,2]
-; AVX2-FAST-NEXT:    vmovaps 96(%rsi), %ymm1
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm1[1,1,2,2,5,5,6,6]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm0[1,2],ymm15[3,4],ymm0[5,6],ymm15[7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm14[1],ymm0[2,3,4],ymm14[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm0[2,2,2,2]
+; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovaps 96(%rsi), %ymm0
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm1[1,2],ymm15[3,4],ymm1[5,6],ymm15[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1,2],ymm12[3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0,1,2],ymm11[3,4],ymm15[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 112(%r8), %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0],ymm12[1,2,3,4],ymm15[5],ymm12[6,7]
-; AVX2-FAST-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm12 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0],ymm11[1,2,3,4],ymm15[5],ymm11[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm2[0,2,3,3,4,6,7,7]
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm14[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0],ymm12[1,2],ymm15[3,4],ymm12[5,6],ymm15[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0],ymm11[1,2],ymm15[3,4],ymm11[5,6],ymm15[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[2,1,3,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm15[3,4],ymm12[5,6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,1,3,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm15[3,4],ymm11[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 120(%r8), %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm15[2],ymm12[3,4,5,6],ymm15[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm13[0,1,2,3],ymm9[4],ymm13[5,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2],ymm11[3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4],ymm12[5,6,7]
 ; AVX2-FAST-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm13 = mem[0,1,3,0,4,5,7,4]
 ; AVX2-FAST-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1],ymm13[2],ymm15[3,4,5,6],ymm13[7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm13[1,2,3],ymm9[4,5],ymm13[6,7]
-; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm9 = ymm9[0,1,2],mem[3],ymm9[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3],ymm5[4],ymm8[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm10[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm11[2,3,0,1,6,7,4,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1],ymm8[2],ymm10[3,4,5,6],ymm8[7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm8[1,2,3],ymm5[4,5],ymm8[6,7]
-; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm5 = ymm5[0,1,2],mem[3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4],ymm4[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm6[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[2,3,0,1,6,7,4,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1],ymm4[2],ymm6[3,4,5,6],ymm4[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1,2,3],ymm12[4,5],ymm13[6,7]
+; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm12 = ymm12[0,1,2],mem[3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4],ymm7[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm9[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm10[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2],ymm9[3,4,5,6],ymm8[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1,2,3],ymm7[4,5],ymm8[6,7]
+; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4],ymm3[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm5[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm6[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3,4,5,6],ymm4[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4,5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,1,3,0,4,5,7,4]
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm14[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6],ymm1[7]
@@ -2733,14 +2718,14 @@ define void @store_i32_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, 544(%r9)
 ; AVX2-FAST-NEXT:    vmovaps %ymm3, 384(%r9)
-; AVX2-FAST-NEXT:    vmovaps %ymm5, 224(%r9)
-; AVX2-FAST-NEXT:    vmovaps %ymm9, 64(%r9)
-; AVX2-FAST-NEXT:    vmovaps %ymm12, 608(%r9)
+; AVX2-FAST-NEXT:    vmovaps %ymm7, 224(%r9)
+; AVX2-FAST-NEXT:    vmovaps %ymm12, 64(%r9)
+; AVX2-FAST-NEXT:    vmovaps %ymm11, 608(%r9)
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, 576(%r9)
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, 480(%r9)
-; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, 448(%r9)
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, 416(%r9)
@@ -2766,92 +2751,89 @@ define void @store_i32_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; 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, 32(%r9)
-; AVX2-FAST-NEXT:    addq $568, %rsp # imm = 0x238
+; AVX2-FAST-NEXT:    addq $584, %rsp # imm = 0x248
 ; AVX2-FAST-NEXT:    vzeroupper
 ; AVX2-FAST-NEXT:    retq
 ;
 ; AVX2-FAST-PERLANE-LABEL: store_i32_stride5_vf32:
 ; AVX2-FAST-PERLANE:       # %bb.0:
 ; AVX2-FAST-PERLANE-NEXT:    subq $552, %rsp # imm = 0x228
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %ymm13
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %ymm15
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r8), %ymm14
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r8), %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r8), %ymm14
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r8), %ymm11
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%r8), %ymm10
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %xmm4
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%r8), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %xmm7
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %xmm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rcx), %xmm0
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %xmm5
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %xmm8
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdx), %xmm1
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm7
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm8
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0,1],ymm6[2,3],ymm9[4,5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, (%r8), %ymm15, %ymm9
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0],ymm6[1,2,3],ymm9[4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm9
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %xmm6
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm7 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %xmm8
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm10
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm12[0,1],ymm5[2,3],ymm12[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, (%r8), %ymm11, %ymm11
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm11[0],ymm5[1,2,3],ymm11[4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2],xmm7[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm15[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3,4,5],ymm8[6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,1,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3],ymm4[4,5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm14[0,1,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4,5],ymm5[6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm8[0],xmm6[0],xmm8[1],xmm6[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 32(%r8), %ymm13, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2,3],ymm5[4],ymm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %xmm4
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,1,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm5[2,3],ymm2[4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm11[0,1,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3,4,5],ymm3[6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 64(%r8), %ymm12, %ymm3
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3],ymm3[4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm6[2],xmm4[2],xmm6[3],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm12[0],xmm6[0],xmm12[1],xmm6[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm8[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 32(%r8), %ymm8, %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1,2,3],ymm8[4],ymm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm12[2],xmm6[2],xmm12[3],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3],ymm3[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm14[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4,5],ymm4[6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm4[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3],ymm6[4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 64(%r8), %ymm4, %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3],ymm4[4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm10[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm13[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rcx), %xmm0
@@ -2863,158 +2845,161 @@ define void @store_i32_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %ymm14
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 96(%r8), %ymm14, %ymm5
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 96(%r8), %ymm5, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %ymm14
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%r8), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%r8), %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm3[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %ymm11
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %ymm10
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm11[1],ymm0[2,3,4],ymm11[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3,4],ymm5[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %ymm9
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[1,1,2,2,5,5,6,6]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1,2],ymm1[3,4],ymm15[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2],ymm1[3,4],ymm6[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 16(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm10[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm11[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm14[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm15[2],ymm9[3],ymm15[3],ymm9[6],ymm15[6],ymm9[7],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm13[2],ymm6[2],ymm13[3],ymm6[3],ymm13[6],ymm6[6],ymm13[7],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 24(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %ymm8
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %ymm7
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1],ymm0[2,3,4],ymm8[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %ymm11
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %ymm6
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[1,1,2,2,5,5,6,6]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1,2],ymm1[3,4],ymm13[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %ymm9
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1,2],ymm1[3,4],ymm10[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 48(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm13[2],ymm6[3],ymm13[3],ymm6[6],ymm13[6],ymm6[7],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[6],ymm10[6],ymm9[7],ymm10[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 56(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdx), %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rcx), %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm4[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3,4],ymm5[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdx), %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rcx), %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1],ymm0[2,3,4],ymm8[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %ymm3
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[1,1,2,2,5,5,6,6]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm12[1,2],ymm1[3,4],ymm12[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %ymm5
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2],ymm1[3,4],ymm6[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 80(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm4[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm12[2],ymm3[3],ymm12[3],ymm3[6],ymm12[6],ymm3[7],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm6[2],ymm5[3],ymm6[3],ymm5[6],ymm6[6],ymm5[7],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 88(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdx), %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rcx), %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4],ymm2[5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rsi), %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm0[1,1,2,2,5,5,6,6]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1,2],ymm13[3,4],ymm14[5,6],ymm13[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4],ymm13[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 112(%r8), %ymm13
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm12[1,2,3,4],ymm13[5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm12 = ymm1[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm2[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0],ymm12[1,2],ymm15[3,4],ymm12[5,6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm0[2],ymm14[2],ymm0[3],ymm14[3],ymm0[6],ymm14[6],ymm0[7],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdx), %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rcx), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rsi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm2[1,2],ymm15[3,4],ymm2[5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2],ymm0[3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 112(%r8), %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0],ymm0[1,2,3,4],ymm15[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0],ymm0[1,2],ymm15[3,4],ymm0[5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[2,1,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm15[3,4],ymm12[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4],ymm0[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 120(%r8), %ymm15
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm15[2],ymm12[3,4,5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0,1],ymm15[2],ymm0[3,4,5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm13[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm14[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1],ymm13[2],ymm14[3,4,5,6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1,2,3],ymm0[4,5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm9 = ymm9[0,1,2,3],mem[4],ymm9[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4],ymm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm11[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2],ymm11[3,4,5,6],ymm10[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4,5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm9 = ymm9[0,1,2],mem[3],ymm9[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm6 = ymm6[0,1,2,3],mem[4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2],ymm8[3,4,5,6],ymm7[7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4,5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm6 = ymm6[0,1,2],mem[3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm3 = ymm3[0,1,2,3],mem[4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,3,0,1,6,7,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3,4,5,6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4,5],ymm4[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm14[4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 544(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm3, 384(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm6, 224(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 64(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm12, 608(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm13, 576(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm8[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2],ymm7[3,4,5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm5[0,1,2],mem[3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm1[0,1,2],mem[3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, 544(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm5, 384(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 224(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 64(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm15, 608(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 576(%r9)
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 448(%r9)
 ; AVX2-FAST-PERLANE-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
@@ -4109,209 +4094,196 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride5_vf64:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $1944, %rsp # imm = 0x798
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm9
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm5
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm11
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm11[1],xmm2[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm11[0],xmm2[0],xmm11[1],xmm2[1]
+; AVX1-ONLY-NEXT:    subq $1784, %rsp # imm = 0x6F8
+; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm12
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm7[1],xmm8[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm3 = xmm7[0],xmm8[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm0[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,0]
+; AVX1-ONLY-NEXT:    vbroadcastss 4(%rdx), %xmm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm4, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3],ymm1[4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r8), %ymm3, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm1[1,2,3],ymm3[4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm14[1],xmm12[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm3 = xmm14[0],xmm12[0],zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm7
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm7[0],xmm8[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,0]
-; AVX1-ONLY-NEXT:    vbroadcastss 4(%rdx), %xmm4
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3],ymm1[4,5,6],ymm3[7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r8), %ymm5, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0],ymm1[1,2,3],ymm3[4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm4
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm3 = zero,xmm1[1],xmm4[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm5, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,2,0]
-; AVX1-ONLY-NEXT:    vbroadcastss 36(%rdx), %xmm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm5, %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3],ymm3[4,5,6],ymm5[7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r8), %ymm9, %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0],ymm3[1,2,3],ymm5[4],ymm3[5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm13
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm12
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm3 = zero,xmm12[1],xmm13[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm12[0],xmm13[0],xmm12[1],xmm13[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm5, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm2[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,0]
+; AVX1-ONLY-NEXT:    vbroadcastss 36(%rdx), %xmm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm4, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3],ymm1[4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r8), %ymm3, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm1[1,2,3],ymm3[4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm13
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,2,0]
-; AVX1-ONLY-NEXT:    vbroadcastss 68(%rdx), %xmm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm5, %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3],ymm3[4,5,6],ymm5[7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm5
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r8), %ymm5, %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0],ymm3[1,2,3],ymm5[4],ymm3[5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm5 = zero,xmm3[1],xmm0[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm6, %ymm5
+; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vbroadcastss 68(%rdx), %xmm3
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm3 = zero,xmm13[1],xmm10[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm4 = xmm13[0],xmm10[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm4, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r8), %ymm4, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm1[1,2,3],ymm3[4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm6
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm6[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,0]
-; AVX1-ONLY-NEXT:    vbroadcastss 100(%rdx), %xmm14
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3],ymm5[4,5,6],ymm6[7]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vbroadcastss 100(%rdx), %xmm3
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm11
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm4 = zero,xmm5[1],xmm11[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm6 = xmm5[0],xmm11[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm6, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3],ymm4[4,5,6],ymm3[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r8), %ymm6, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0],ymm3[1,2,3],ymm4[4],ymm3[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r8), %ymm0, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0],ymm5[1,2,3],ymm6[4],ymm5[5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm5
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm6 = zero,xmm5[1],xmm0[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm14, %ymm6
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rcx), %xmm9
-; AVX1-ONLY-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm9[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[0,1,2,0]
-; AVX1-ONLY-NEXT:    vbroadcastss 132(%rdx), %xmm15
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm14[2,3],ymm6[4,5,6],ymm14[7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%r8), %ymm0, %ymm14
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm14[0],ymm6[1,2,3],ymm14[4],ymm6[5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 128(%rcx), %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:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; AVX1-ONLY-NEXT:    vbroadcastss 132(%rdx), %xmm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm4
+; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm6
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm14 = zero,xmm6[1],xmm0[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm15, %ymm14
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm6 = zero,xmm9[1],xmm0[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm15 = xmm9[0],xmm0[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm15, %ymm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1],ymm4[2,3],ymm6[4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%r8), %ymm15, %ymm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0],ymm4[1,2,3],ymm6[4],ymm4[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %xmm9
-; AVX1-ONLY-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm9[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[0,1,2,0]
-; AVX1-ONLY-NEXT:    vbroadcastss 164(%rdx), %xmm9
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm15, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm14[0,1],ymm9[2,3],ymm14[4,5,6],ymm9[7]
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm10
-; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%r8), %ymm10, %ymm14
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm14[0],ymm9[1,2,3],ymm14[4],ymm9[5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %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:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,0]
+; AVX1-ONLY-NEXT:    vbroadcastss 164(%rdx), %xmm6
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm4, %ymm6
+; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm15 = zero,xmm4[1],xmm0[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm0 = xmm4[0],xmm0[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm15[0,1],ymm6[2,3],ymm15[4,5,6],ymm6[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%r8), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm6[1,2,3],ymm0[4],ymm6[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm14
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm9 = zero,xmm14[1],xmm0[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm15, %ymm9
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; AVX1-ONLY-NEXT:    vbroadcastss 196(%rdx), %xmm6
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm10
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm10[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[0,1,2,0]
-; AVX1-ONLY-NEXT:    vbroadcastss 196(%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm15, %ymm10
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm10[2,3],ymm9[4,5,6],ymm10[7]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm15 = zero,xmm6[1],xmm0[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm0 = xmm6[0],xmm0[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0,1],ymm1[2,3],ymm15[4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%r8), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%r8), %ymm0, %ymm10
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0],ymm9[1,2,3],ymm10[4],ymm9[5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm10
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm9 = zero,xmm10[1],xmm0[1],zero
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm10, %ymm9
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; AVX1-ONLY-NEXT:    vbroadcastss 228(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %xmm10
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm10 = xmm10[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,2,0]
-; AVX1-ONLY-NEXT:    vbroadcastss 228(%rdx), %xmm15
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm10, %ymm10
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm10[2,3],ymm9[4,5,6],ymm10[7]
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm10
-; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%r8), %ymm10, %ymm10
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0],ymm9[1,2,3],ymm10[4],ymm9[5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm7 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
-; AVX1-ONLY-NEXT:    vbroadcastss 4(%rcx), %xmm8
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm8, %ymm7
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,zero,xmm11[2],xmm2[2]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm0[2,3],ymm7[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4,5],ymm2[6],ymm0[7]
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm0[1],xmm1[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm15 = xmm0[0],xmm1[0],zero,zero
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm15, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm3[2,3],ymm1[4,5,6],ymm3[7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%r8), %ymm15, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4],ymm0[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, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX1-ONLY-NEXT:    vbroadcastss 4(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,zero,xmm7[2],xmm8[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rsi), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm7 = zero,zero,xmm2[2],xmm0[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[1,1],xmm0[1,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,zero,xmm1[2],xmm0[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,1],xmm0[1,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm15
+; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm15[1,1],ymm1[5,5],ymm15[5,5]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm2
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm2[3,3],xmm1[3,3]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,3],ymm2[3,3],ymm3[7,7],ymm2[7,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm7
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[1,1],ymm2[1,1],ymm7[5,5],ymm2[5,5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm7
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm7[3,3],xmm2[3,3]
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm7[2],xmm2[2],xmm7[3],xmm2[3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm2, %ymm2
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm7
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm7 = ymm7[3,3],mem[3,3],ymm7[7,7],mem[7,7]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm7[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm7[3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm7
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm7[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0],ymm0[1,2,3,4],ymm7[5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4],ymm2[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm7[2],ymm2[3,4,5,6],ymm7[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm2[2],ymm1[3,4,5,6],ymm2[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, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vbroadcastss 36(%rcx), %xmm2
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,zero,xmm1[2],xmm4[2]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vbroadcastss 36(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,zero,xmm14[2],xmm12[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
@@ -4323,20 +4295,20 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %ymm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm1[1,1],ymm2[5,5],ymm1[5,5]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %ymm14
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm14[1,1],ymm1[1,1],ymm14[5,5],ymm1[5,5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm2[3,3],xmm1[3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm2[3,3],xmm1[3,3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = ymm2[3,3],mem[3,3],ymm2[7,7],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,3],ymm2[3,3],ymm3[7,7],ymm2[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm2
@@ -4351,8 +4323,8 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX1-ONLY-NEXT:    vbroadcastss 68(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,zero,xmm12[2],xmm13[2]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,zero,xmm13[2],xmm10[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
@@ -4362,22 +4334,22 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,zero,xmm1[2],xmm0[2]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,1],xmm0[1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm13
+; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %ymm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm1[1,1],ymm2[5,5],ymm1[5,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm13[1,1],ymm1[5,5],ymm13[5,5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm2[3,3],xmm1[3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm2[3,3],xmm1[3,3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = ymm2[3,3],mem[3,3],ymm2[7,7],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,3],ymm2[3,3],ymm3[7,7],ymm2[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm2
@@ -4392,9 +4364,8 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX1-ONLY-NEXT:    vbroadcastss 100(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = zero,zero,xmm3[2],mem[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm1 = zero,zero,xmm5[2],xmm11[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
@@ -4413,13 +4384,14 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 112(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps 112(%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[3,3],xmm1[3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm2[3,3],xmm1[3,3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 96(%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 {{.*#+}} ymm2 = ymm2[3,3],ymm15[3,3],ymm2[7,7],ymm15[7,7]
+; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,3],ymm2[3,3],ymm3[7,7],ymm2[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %ymm2
@@ -4434,9 +4406,9 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX1-ONLY-NEXT:    vbroadcastss 132(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = zero,zero,xmm5[2],mem[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = zero,zero,xmm9[2],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
@@ -4458,10 +4430,11 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[3,3],xmm1[3,3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-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 {{.*#+}} ymm2 = ymm2[3,3],ymm13[3,3],ymm2[7,7],ymm13[7,7]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,3],ymm2[3,3],ymm3[7,7],ymm2[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %ymm2
@@ -4476,9 +4449,9 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX1-ONLY-NEXT:    vbroadcastss 164(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = zero,zero,xmm6[2],mem[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = zero,zero,xmm4[2],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
@@ -4488,11 +4461,11 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,zero,xmm1[2],xmm0[2]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,1],xmm0[1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm1[1,1],ymm2[5,5],ymm1[5,5]
+; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm2[1,1],ymm1[5,5],ymm2[5,5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 176(%rcx), %xmm1
@@ -4500,26 +4473,27 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[3,3],xmm1[3,3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %ymm12
-; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = ymm12[3,3],mem[3,3],ymm12[7,7],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[3,3],ymm12[3,3],ymm2[7,7],ymm12[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4],ymm2[5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm2[2],ymm1[3,4,5,6],ymm2[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm2[2],ymm1[3,4,5,6],ymm2[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, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX1-ONLY-NEXT:    vbroadcastss 196(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = zero,zero,xmm14[2],mem[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = zero,zero,xmm6[2],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
@@ -4531,8 +4505,8 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %ymm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %ymm10
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[1,1],ymm1[1,1],ymm10[5,5],ymm1[5,5]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %ymm11
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm11[1,1],ymm1[1,1],ymm11[5,5],ymm1[5,5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 208(%rcx), %xmm1
@@ -4540,13 +4514,13 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[3,3],xmm1[3,3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %ymm6
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm6[3,3],ymm9[3,3],ymm6[7,7],ymm9[7,7]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[3,3],ymm8[3,3],ymm7[7,7],ymm8[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %ymm11
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm11[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %ymm9
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm9[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4],ymm2[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm2[2],ymm1[3,4,5,6],ymm2[7]
@@ -4569,78 +4543,77 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,zero,xmm1[2],xmm0[2]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,1],xmm0[1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %ymm7
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %ymm5
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %ymm4
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[1,1],ymm7[1,1],ymm4[5,5],ymm7[5,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[1,1],ymm5[1,1],ymm4[5,5],ymm5[5,5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 240(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 240(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm1[3,3],xmm0[3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3],xmm0[3,3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[3,3],ymm3[3,3],ymm0[7,7],ymm3[7,7]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm1[0,1,2],ymm5[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 224(%r8), %ymm1
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0],ymm2[1,2,3,4],ymm8[5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm8[2],ymm5[3,4,5,6],ymm8[7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = mem[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = mem[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm8 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2],ymm8[3,4,5,6],ymm5[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1,2,3],ymm2[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = mem[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = mem[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm8 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2],ymm8[3,4,5,6],ymm5[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1,2,3],ymm2[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm8 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = mem[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = mem[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm0[3,3],ymm1[3,3],ymm0[7,7],ymm1[7,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm6[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm2[0,1,2],ymm6[3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 224(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm10[0],ymm3[1,2,3,4],ymm10[5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm10[2],ymm6[3,4,5,6],ymm10[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = mem[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[0,1,2,3],mem[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = mem[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm10 = ymm15[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm6[2],ymm10[3,4,5,6],ymm6[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm6[1,2,3],ymm3[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = mem[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[0,1,2,3],mem[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm14[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm10 = mem[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm6[2],ymm10[3,4,5,6],ymm6[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm6[1,2,3],ymm3[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm10 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = mem[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[0,1,2,3],mem[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = mem[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm15 = ymm13[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm15[0,1],ymm6[2],ymm15[3,4,5,6],ymm6[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm6[1,2,3],ymm3[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = mem[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[0,1,2,3],mem[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = mem[0,1,3,0,4,5,7,4]
 ; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm14[0,1],ymm5[2],ymm14[3,4,5,6],ymm5[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1,2,3],ymm2[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = mem[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm15[4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm14 = mem[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm15 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4,5,6],ymm14[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm14[1,2,3],ymm2[4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3,4,5,6],ymm15[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm14[1,2,3],ymm3[4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = mem[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm13[4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = ymm14[0,1,2,3],mem[4],ymm14[5,6,7]
 ; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm15 = mem[0,1,3,0,4,5,7,4]
 ; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
@@ -4649,39 +4622,39 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0],ymm13[1,2,3],ymm14[4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm13 = ymm13[0,1,2],mem[3],ymm13[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm12 = ymm12[0,1,2,3],mem[4],ymm12[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm14 = mem[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm15 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4,5,6],ymm14[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm14[1,2,3],ymm12[4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = mem[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm12[4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = mem[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm12 = mem[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm15[2],ymm12[3,4,5,6],ymm15[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm14[0],ymm12[1,2,3],ymm14[4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm12 = ymm12[0,1,2],mem[3],ymm12[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm9[4],ymm6[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm10[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm10 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1],ymm9[2],ymm10[3,4,5,6],ymm9[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm9[1,2,3],ymm6[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm11[3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4],ymm7[5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm11[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm11 = mem[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0,1],ymm8[2],ymm11[3,4,5,6],ymm8[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1,2,3],ymm7[4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3],ymm7[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm4 = ymm7[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3,4,5,6],ymm3[7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1,2,3],ymm0[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm4 = ymm5[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2],ymm4[3,4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1184(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm6, 1024(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 1024(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm12, 864(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm13, 704(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm2, 544(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm5, 384(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm8, 224(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 544(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm6, 384(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %ymm10, 224(%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
@@ -4696,9 +4669,9 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1056(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 992(%r9)
-; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 928(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 928(%r9)
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 896(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 832(%r9)
@@ -4748,93 +4721,89 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; 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, (%r9)
-; AVX1-ONLY-NEXT:    addq $1944, %rsp # imm = 0x798
+; AVX1-ONLY-NEXT:    addq $1784, %rsp # imm = 0x6F8
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: store_i32_stride5_vf64:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    subq $1768, %rsp # imm = 0x6E8
-; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %ymm14
-; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %ymm11
-; AVX2-SLOW-NEXT:    vmovaps (%rdi), %ymm10
+; AVX2-SLOW-NEXT:    subq $1736, %rsp # imm = 0x6C8
 ; AVX2-SLOW-NEXT:    vmovaps (%r8), %ymm15
 ; AVX2-SLOW-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 32(%r8), %ymm13
+; AVX2-SLOW-NEXT:    vmovaps 32(%r8), %ymm14
+; AVX2-SLOW-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 64(%r8), %ymm13
 ; AVX2-SLOW-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 64(%r8), %ymm12
-; AVX2-SLOW-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps (%rcx), %xmm4
-; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %xmm2
+; AVX2-SLOW-NEXT:    vmovaps (%rcx), %xmm7
+; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %xmm3
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rcx), %xmm0
-; AVX2-SLOW-NEXT:    vmovaps (%rdx), %xmm5
-; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %xmm3
+; AVX2-SLOW-NEXT:    vmovaps (%rdx), %xmm8
+; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %xmm4
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rdx), %xmm1
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
-; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm7
-; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm8
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0,1],ymm6[2,3],ymm9[4,5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, (%r8), %ymm10, %ymm9
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0],ymm6[1,2,3],ymm9[4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[0,0,2,1]
+; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm9
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %xmm6
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm7 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
-; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %xmm8
+; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %xmm2
+; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm10
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm12[0,1],ymm5[2,3],ymm12[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %xmm12
+; AVX2-SLOW-NEXT:    vinsertf128 $1, (%r8), %ymm11, %ymm11
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm11[0],ymm5[1,2,3],ymm11[4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %xmm5
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,2,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,0,2,1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[1,1,2,2]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2],xmm7[3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,1,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm15[0,1,2,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3,4,5],ymm8[6],ymm7[7]
+; AVX2-SLOW-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,1,2,2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,1,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3],ymm4[4,5],ymm7[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm15[0,1,2,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4,5],ymm5[6],ymm4[7]
-; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,0,2,1]
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm8[0],xmm6[0],xmm8[1],xmm6[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6],ymm4[7]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, 32(%r8), %ymm11, %ymm5
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2,3],ymm5[4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %xmm4
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %xmm6
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,2,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,1,2,2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2],xmm2[3]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,1,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm5[2,3],ymm2[4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm13[0,1,2,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3,4,5],ymm3[6],ymm2[7]
-; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, 64(%r8), %ymm14, %ymm3
-; AVX2-SLOW-NEXT:    vmovaps %ymm14, %ymm9
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3],ymm3[4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm6[2],xmm4[2],xmm6[3],xmm4[3]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm12[0],xmm6[0],xmm12[1],xmm6[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm8[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6],ymm7[7]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, 32(%r8), %ymm8, %ymm8
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1,2,3],ymm8[4],ymm7[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm12[2],xmm6[2],xmm12[3],xmm6[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,1,2,2]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3],ymm3[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm14[0,1,2,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4,5],ymm4[6],ymm3[7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm4[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3],ymm6[4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, 64(%r8), %ymm4, %ymm4
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3],ymm4[4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm12[0,1,2,1]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm13[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rcx), %xmm0
@@ -4846,22 +4815,21 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %ymm6
-; AVX2-SLOW-NEXT:    vinsertf128 $1, 96(%r8), %ymm6, %ymm5
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vmovaps 96(%r8), %ymm6
+; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vinsertf128 $1, 96(%r8), %ymm5, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vmovaps 96(%r8), %ymm1
-; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rcx), %xmm0
@@ -4873,22 +4841,21 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT:    vmovaps 128(%rdi), %ymm7
-; AVX2-SLOW-NEXT:    vinsertf128 $1, 128(%r8), %ymm7, %ymm5
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, 128(%r8), %ymm5, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
+; AVX2-SLOW-NEXT:    vmovaps 128(%r8), %ymm3
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vmovaps 128(%r8), %ymm1
-; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm3[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rcx), %xmm0
@@ -4900,11 +4867,9 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT:    vmovaps 160(%rdi), %ymm8
-; AVX2-SLOW-NEXT:    vinsertf128 $1, 160(%r8), %ymm8, %ymm5
-; AVX2-SLOW-NEXT:    vmovaps %ymm8, %ymm15
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, 160(%r8), %ymm5, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
@@ -4928,11 +4893,9 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT:    vmovaps 192(%rdi), %ymm8
-; AVX2-SLOW-NEXT:    vinsertf128 $1, 192(%r8), %ymm8, %ymm5
-; AVX2-SLOW-NEXT:    vmovaps %ymm8, %ymm13
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, 192(%r8), %ymm5, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
@@ -4956,10 +4919,8 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT:    vmovaps 224(%rdi), %ymm5
-; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, 224(%r8), %ymm5, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4981,22 +4942,22 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps %ymm1, %ymm2
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm14[1],ymm0[2,3,4],ymm14[5],ymm0[6,7]
-; AVX2-SLOW-NEXT:    vmovaps %ymm14, %ymm3
-; AVX2-SLOW-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovaps (%rsi), %ymm14
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm14[1,1,2,2,5,5,6,6]
-; AVX2-SLOW-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1,2],ymm1[3,4],ymm10[5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vmovaps (%rdi), %ymm3
+; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps (%rsi), %ymm4
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1,2],ymm1[3,4],ymm3[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 16(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm14[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm14[2],ymm10[2],ymm14[3],ymm10[3],ymm14[6],ymm10[6],ymm14[7],ymm10[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[6],ymm3[6],ymm4[7],ymm3[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
@@ -5012,11 +4973,12 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps %ymm1, %ymm2
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %ymm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %ymm4
 ; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1,2],ymm1[3,4],ymm11[5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %ymm5
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 48(%r8), %ymm1
@@ -5025,7 +4987,7 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm11[2],ymm4[3],ymm11[3],ymm4[6],ymm11[6],ymm4[7],ymm11[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
@@ -5041,11 +5003,12 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps %ymm1, %ymm2
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %ymm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %ymm4
 ; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1,2],ymm1[3,4],ymm9[5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %ymm5
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 80(%r8), %ymm1
@@ -5054,7 +5017,7 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm9[2],ymm4[3],ymm9[3],ymm4[6],ymm9[6],ymm4[7],ymm9[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
@@ -5070,10 +5033,12 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps %ymm1, %ymm2
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovaps 96(%rsi), %ymm12
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[1,1,2,2,5,5,6,6]
-; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2],ymm1[3,4],ymm6[5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %ymm4
+; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 96(%rsi), %ymm5
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 112(%r8), %ymm1
@@ -5082,7 +5047,7 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm12[2],ymm6[2],ymm12[3],ymm6[3],ymm12[6],ymm6[6],ymm12[7],ymm6[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
@@ -5090,201 +5055,205 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rdx), %ymm1
-; AVX2-SLOW-NEXT:    vmovaps 128(%rcx), %ymm11
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vmovaps 128(%rcx), %ymm2
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vmovaps %ymm2, %ymm3
+; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-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 %ymm1, %ymm2
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovaps 128(%rsi), %ymm10
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm10[1,1,2,2,5,5,6,6]
-; AVX2-SLOW-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm7[1,2],ymm1[3,4],ymm7[5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vmovaps 128(%rdi), %ymm4
+; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 128(%rsi), %ymm13
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 144(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm10[2],ymm7[2],ymm10[3],ymm7[3],ymm10[6],ymm7[6],ymm10[7],ymm7[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm13[2],ymm4[2],ymm13[3],ymm4[3],ymm13[6],ymm4[6],ymm13[7],ymm4[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 152(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 160(%rdx), %ymm9
-; AVX2-SLOW-NEXT:    vmovaps 160(%rcx), %ymm8
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm9[1],ymm0[2,3,4],ymm9[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 160(%rdx), %ymm12
+; AVX2-SLOW-NEXT:    vmovaps 160(%rcx), %ymm11
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovaps 160(%rsi), %ymm7
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[1,1,2,2,5,5,6,6]
-; AVX2-SLOW-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1,2],ymm1[3,4],ymm15[5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vmovaps 160(%rdi), %ymm9
+; AVX2-SLOW-NEXT:    vmovaps 160(%rsi), %ymm8
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1,2],ymm1[3,4],ymm9[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 176(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm8[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm7[2],ymm15[2],ymm7[3],ymm15[3],ymm7[6],ymm15[6],ymm7[7],ymm15[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 184(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 192(%rdx), %ymm6
-; AVX2-SLOW-NEXT:    vmovaps 192(%rcx), %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm6[1],ymm0[2,3,4],ymm6[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vmovaps 192(%rdx), %ymm7
+; AVX2-SLOW-NEXT:    vmovaps 192(%rcx), %ymm6
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3,4],ymm7[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
+; AVX2-SLOW-NEXT:    vmovaps 192(%rdi), %ymm5
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rsi), %ymm4
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
-; AVX2-SLOW-NEXT:    vmovups %ymm13, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1,2],ymm1[3,4],ymm13[5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1,2],ymm1[3,4],ymm5[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 208(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm13[2],ymm4[3],ymm13[3],ymm4[6],ymm13[6],ymm4[7],ymm13[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 216(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 224(%rdx), %ymm13
-; AVX2-SLOW-NEXT:    vmovaps 224(%rcx), %ymm3
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1],ymm0[2,3,4],ymm13[5],ymm0[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovaps 224(%rsi), %ymm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm2[1,1,2,2,5,5,6,6]
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm0[1,2],ymm15[3,4],ymm0[5,6],ymm15[7]
+; AVX2-SLOW-NEXT:    vmovaps 224(%rdx), %ymm3
+; AVX2-SLOW-NEXT:    vmovaps 224(%rcx), %ymm2
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4],ymm3[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm0[2,2,2,2]
+; AVX2-SLOW-NEXT:    vmovaps 224(%rdi), %ymm1
+; AVX2-SLOW-NEXT:    vmovaps 224(%rsi), %ymm0
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm1[1,2],ymm15[3,4],ymm1[5,6],ymm15[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0,1,2],ymm1[3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0,1,2],ymm10[3,4],ymm15[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 240(%r8), %ymm15
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0],ymm1[1,2,3,4],ymm15[5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm13[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0],ymm1[1,2],ymm15[3,4],ymm1[5,6],ymm15[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[6],ymm0[6],ymm2[7],ymm0[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0],ymm10[1,2,3,4],ymm15[5],ymm10[6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm10, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0],ymm10[1,2],ymm15[3,4],ymm10[5,6],ymm15[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm1[2,1,3,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,1,3,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm15[3,4],ymm10[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 248(%r8), %ymm15
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm15[2],ymm0[3,4,5,6],ymm15[7]
-; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm14[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm15[2],ymm10[3,4,5,6],ymm15[7]
+; AVX2-SLOW-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = mem[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = ymm10[0,1,2,3],mem[4],ymm10[5,6,7]
 ; AVX2-SLOW-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm15 = mem[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm14 = mem[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3,4,5,6],ymm15[7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm14[1,2,3],ymm0[4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = mem[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm14[1,2,3],ymm10[4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = ymm10[0,1,2],mem[3],ymm10[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = mem[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = ymm10[0,1,2,3],mem[4],ymm10[5,6,7]
 ; AVX2-SLOW-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm14 = mem[0,1,3,0,4,5,7,4]
 ; AVX2-SLOW-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4,5,6],ymm14[7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm14[1,2,3],ymm0[4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = mem[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm14[1,2,3],ymm10[4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = ymm10[0,1,2],mem[3],ymm10[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = mem[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = ymm10[0,1,2,3],mem[4],ymm10[5,6,7]
 ; AVX2-SLOW-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm14 = mem[0,1,3,0,4,5,7,4]
 ; AVX2-SLOW-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4,5,6],ymm14[7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm14[1,2,3],ymm0[4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm15 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm12[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm12 = mem[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm14 = mem[2,3,0,1,6,7,4,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm14[0,1],ymm12[2],ymm14[3,4,5,6],ymm12[7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1,2,3],ymm0[4,5],ymm12[6,7]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm12 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm10[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm11[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm11 = mem[2,3,0,1,6,7,4,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2],ymm11[3,4,5,6],ymm10[7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm10[1,2,3],ymm0[4,5],ymm10[6,7]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm7 = ymm7[0,1,2,3],mem[4],ymm7[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,3,0,1,6,7,4,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2],ymm9[3,4,5,6],ymm8[7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1,2,3],ymm7[4,5],ymm8[6,7]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm14[1,2,3],ymm10[4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = ymm10[0,1,2],mem[3],ymm10[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = mem[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = ymm10[0,1,2,3],mem[4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = mem[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4,5,6],ymm14[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm14[1,2,3],ymm10[4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = ymm10[0,1,2],mem[3],ymm10[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm13 = ymm13[0,1,2,3],mem[4],ymm13[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm14 = mem[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4,5,6],ymm14[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1,2,3],ymm13[4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm13 = ymm13[0,1,2],mem[3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm11[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1],ymm9[2],ymm11[3,4,5,6],ymm9[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4,5],ymm9[6,7]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm8 = ymm8[0,1,2],mem[3],ymm8[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps $16, (%rsp), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm4 = ymm4[0,1,2,3],mem[4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm6[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2],ymm6[3,4,5,6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4,5],ymm5[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[3,0,2,3,7,4,6,7]
-; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm13[2,3,0,1,6,7,4,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm1 = ymm1[0,1,2],mem[3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT:    vmovaps %ymm1, 1184(%r9)
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 1184(%r9)
 ; AVX2-SLOW-NEXT:    vmovaps %ymm4, 1024(%r9)
-; AVX2-SLOW-NEXT:    vmovaps %ymm7, 864(%r9)
-; AVX2-SLOW-NEXT:    vmovaps %ymm0, 704(%r9)
-; AVX2-SLOW-NEXT:    vmovaps %ymm12, 544(%r9)
-; AVX2-SLOW-NEXT:    vmovaps %ymm15, 384(%r9)
+; AVX2-SLOW-NEXT:    vmovaps %ymm8, 864(%r9)
+; AVX2-SLOW-NEXT:    vmovaps %ymm13, 704(%r9)
+; AVX2-SLOW-NEXT:    vmovaps %ymm10, 544(%r9)
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps %ymm0, 384(%r9)
 ; 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, 64(%r9)
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps %ymm0, 1248(%r9)
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps %ymm0, 1216(%r9)
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps %ymm0, 1088(%r9)
@@ -5346,85 +5315,85 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; 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:    addq $1768, %rsp # imm = 0x6E8
+; AVX2-SLOW-NEXT:    addq $1736, %rsp # imm = 0x6C8
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: store_i32_stride5_vf64:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    subq $1800, %rsp # imm = 0x708
-; AVX2-FAST-NEXT:    vmovaps (%r8), %ymm9
-; AVX2-FAST-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 32(%r8), %ymm8
-; AVX2-FAST-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps (%rsi), %xmm7
-; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %xmm13
-; AVX2-FAST-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 64(%rsi), %xmm1
-; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vmovaps (%rdi), %xmm10
-; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %xmm4
-; AVX2-FAST-NEXT:    vmovaps 64(%rdi), %xmm12
+; AVX2-FAST-NEXT:    vmovaps (%r8), %ymm7
+; AVX2-FAST-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 32(%r8), %ymm6
+; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps (%rsi), %xmm10
+; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %xmm4
+; AVX2-FAST-NEXT:    vmovaps 64(%rsi), %xmm12
 ; AVX2-FAST-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm10[2],xmm7[2],xmm10[3],xmm7[3]
-; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm14 = <0,1,0,1,u,u,2,2>
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
-; AVX2-FAST-NEXT:    vmovaps (%rdx), %xmm11
-; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %xmm5
-; AVX2-FAST-NEXT:    vmovaps (%rcx), %xmm3
-; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %xmm6
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm3[1,1,2,2]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm11[2],xmm2[3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm9[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4,5],ymm2[6],ymm0[7]
+; AVX2-FAST-NEXT:    vmovaps (%rdi), %xmm11
+; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %xmm5
+; AVX2-FAST-NEXT:    vmovaps 64(%rdi), %xmm3
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = <0,1,0,1,u,u,2,2>
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-NEXT:    vmovaps (%rdx), %xmm13
+; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %xmm8
+; AVX2-FAST-NEXT:    vmovaps (%rcx), %xmm14
+; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %xmm9
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm14[1,1,2,2]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2],xmm1[3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm7[0,1,2,1]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm4[2],xmm13[2],xmm4[3],xmm13[3]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm6[1,1,2,2]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm5[2],xmm2[3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm8[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4,5],ymm2[6],ymm0[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm9[1,1,2,2]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm8[2],xmm1[3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm6[0,1,2,1]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm12[2],xmm1[2],xmm12[3],xmm1[3]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm2
-; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %xmm12
-; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %xmm9
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm8 = xmm9[1,1,2,2]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm12[2],xmm8[3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0,1],ymm2[2,3],ymm8[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT:    vmovaps 64(%r8), %ymm8
-; AVX2-FAST-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm8[1],ymm2[2,3,4,5],ymm8[6],ymm2[7]
-; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 96(%rsi), %xmm0
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm3[2],xmm12[2],xmm3[3],xmm12[3]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %xmm6
+; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %xmm7
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[1,1,2,2]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm6[2],xmm1[3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 64(%r8), %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 96(%rsi), %xmm1
+; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %xmm1
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm15
+; AVX2-FAST-NEXT:    vmovaps 96(%rdx), %xmm0
+; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm14, %ymm13
-; AVX2-FAST-NEXT:    vmovaps 96(%rdx), %xmm8
-; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %xmm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm2[1,1,2,2]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm8[2],xmm15[3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1],ymm13[2,3],ymm15[4,5],ymm13[6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm12 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm12 = xmm12[0,1],xmm0[2],xmm12[3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[0,1,1,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm15[2,3],ymm12[4,5],ymm15[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 96(%r8), %ymm0
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm0[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm15[1],ymm13[2,3,4,5],ymm15[6],ymm13[7]
-; AVX2-FAST-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm15[1],ymm12[2,3,4,5],ymm15[6],ymm12[7]
+; AVX2-FAST-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 128(%rsi), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 128(%rdi), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm13 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX2-FAST-NEXT:    vpermps %ymm13, %ymm14, %ymm13
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm12 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX2-FAST-NEXT:    vpermps %ymm12, %ymm2, %ymm12
 ; AVX2-FAST-NEXT:    vmovaps 128(%rdx), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 128(%rcx), %xmm1
@@ -5432,37 +5401,37 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2],xmm15[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1],ymm13[2,3],ymm15[4,5],ymm13[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3],ymm15[4,5],ymm12[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 128(%r8), %ymm0
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm0[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm15[1],ymm13[2,3,4,5],ymm15[6],ymm13[7]
-; AVX2-FAST-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm15[1],ymm12[2,3,4,5],ymm15[6],ymm12[7]
+; AVX2-FAST-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 160(%rsi), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 160(%rdi), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm13 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX2-FAST-NEXT:    vpermps %ymm13, %ymm14, %ymm13
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm12 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX2-FAST-NEXT:    vpermps %ymm12, %ymm2, %ymm12
 ; AVX2-FAST-NEXT:    vmovaps 160(%rdx), %xmm0
-; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 160(%rcx), %xmm1
-; AVX2-FAST-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
+; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2],xmm15[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1],ymm13[2,3],ymm15[4,5],ymm13[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3],ymm15[4,5],ymm12[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 160(%r8), %ymm0
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm0[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm15[1],ymm13[2,3,4,5],ymm15[6],ymm13[7]
-; AVX2-FAST-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm15[1],ymm12[2,3,4,5],ymm15[6],ymm12[7]
+; AVX2-FAST-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 192(%rsi), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 192(%rdi), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm13 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX2-FAST-NEXT:    vpermps %ymm13, %ymm14, %ymm13
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm12 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX2-FAST-NEXT:    vpermps %ymm12, %ymm2, %ymm12
 ; AVX2-FAST-NEXT:    vmovaps 192(%rdx), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 192(%rcx), %xmm1
@@ -5470,18 +5439,18 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2],xmm15[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1],ymm13[2,3],ymm15[4,5],ymm13[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3],ymm15[4,5],ymm12[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 192(%r8), %ymm0
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm0[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm15[1],ymm13[2,3,4,5],ymm15[6],ymm13[7]
-; AVX2-FAST-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm15[1],ymm12[2,3,4,5],ymm15[6],ymm12[7]
+; AVX2-FAST-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 224(%rsi), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 224(%rdi), %xmm13
-; AVX2-FAST-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm13 = xmm13[2],xmm0[2],xmm13[3],xmm0[3]
-; AVX2-FAST-NEXT:    vpermps %ymm13, %ymm14, %ymm13
+; AVX2-FAST-NEXT:    vmovaps 224(%rdi), %xmm12
+; AVX2-FAST-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm12 = xmm12[2],xmm0[2],xmm12[3],xmm0[3]
+; AVX2-FAST-NEXT:    vpermps %ymm12, %ymm2, %ymm12
 ; AVX2-FAST-NEXT:    vmovaps 224(%rdx), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 224(%rcx), %xmm0
@@ -5489,409 +5458,413 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm0[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2],xmm15[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1],ymm13[2,3],ymm15[4,5],ymm13[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3],ymm15[4,5],ymm12[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 224(%r8), %ymm0
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm0[0,1,2,1]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm15[1],ymm13[2,3,4,5],ymm15[6],ymm13[7]
-; AVX2-FAST-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm11[0],xmm3[0],xmm11[1],xmm3[1]
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm10[0],xmm7[0],xmm10[1],xmm7[1]
-; AVX2-FAST-NEXT:    vpermps %ymm11, %ymm14, %ymm10
-; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm15 = [0,1,3,2,3,2,3,2]
-; AVX2-FAST-NEXT:    vpermps %ymm7, %ymm15, %ymm7
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm10[2,3],ymm7[4,5,6],ymm10[7]
-; AVX2-FAST-NEXT:    vmovaps (%rdi), %ymm1
-; AVX2-FAST-NEXT:    vinsertf128 $1, (%r8), %ymm1, %ymm10
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm10[0],ymm7[1,2,3],ymm10[4],ymm7[5,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm15[1],ymm12[2,3,4,5],ymm15[6],ymm12[7]
+; AVX2-FAST-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
+; AVX2-FAST-NEXT:    vpermps %ymm12, %ymm2, %ymm11
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm14 = [0,1,3,2,3,2,3,2]
+; AVX2-FAST-NEXT:    vpermps %ymm10, %ymm14, %ymm12
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3],ymm12[4,5,6],ymm11[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, (%r8), %ymm10, %ymm10
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm10[0],ymm11[1,2,3],ymm10[4],ymm11[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps (%rdx), %ymm13
-; AVX2-FAST-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps (%rcx), %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm0[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vmovaps %ymm0, %ymm3
+; AVX2-FAST-NEXT:    vmovaps (%rdx), %ymm0
+; AVX2-FAST-NEXT:    vmovaps (%rcx), %ymm1
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm1[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm13
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm0[1],ymm10[2,3,4],ymm0[5],ymm10[6,7]
+; AVX2-FAST-NEXT:    vmovaps %ymm0, %ymm12
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm13[1],ymm7[2,3,4],ymm13[5],ymm7[6,7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,2]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,2]
+; AVX2-FAST-NEXT:    vmovaps (%rdi), %ymm15
 ; AVX2-FAST-NEXT:    vmovaps (%rsi), %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm0[1,1,2,2,5,5,6,6]
-; AVX2-FAST-NEXT:    vmovaps %ymm0, %ymm11
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vmovaps %ymm0, %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm1[1,2],ymm10[3,4],ymm1[5,6],ymm10[7]
-; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2],ymm7[3,4],ymm10[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 16(%r8), %ymm10
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm10[0],ymm7[1,2,3,4],ymm10[5],ymm7[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm15[1,2],ymm11[3,4],ymm15[5,6],ymm11[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3,4],ymm11[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 16(%r8), %ymm11
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0],ymm10[1,2,3,4],ymm11[5],ymm10[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm13[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0],ymm7[1,2],ymm10[3,4],ymm7[5,6],ymm10[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm10 = ymm11[2],ymm1[2],ymm11[3],ymm1[3],ymm11[6],ymm1[6],ymm11[7],ymm1[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[2,1,3,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm10[3,4],ymm7[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 24(%r8), %ymm10
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm10[2],ymm7[3,4,5,6],ymm10[7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm13[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0],ymm10[1,2],ymm11[3,4],ymm10[5,6],ymm11[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm11 = ymm1[2],ymm15[2],ymm1[3],ymm15[3],ymm1[6],ymm15[6],ymm1[7],ymm15[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[3,3,3,3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,1,3,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3,4],ymm10[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 24(%r8), %ymm11
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm10[0,1],ymm11[2],ymm10[3,4,5,6],ymm11[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
-; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm3 # 16-byte Folded Reload
-; AVX2-FAST-NEXT:    # xmm3 = xmm4[0],mem[0],xmm4[1],mem[1]
-; AVX2-FAST-NEXT:    vpermps %ymm5, %ymm14, %ymm4
-; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm15, %ymm3
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3],ymm3[4,5,6],ymm4[7]
-; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %ymm1
-; AVX2-FAST-NEXT:    vinsertf128 $1, 32(%r8), %ymm1, %ymm4
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0],ymm3[1,2,3],ymm4[4],ymm3[5,6,7]
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; AVX2-FAST-NEXT:    vpermps %ymm8, %ymm2, %ymm5
+; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm14, %ymm8
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3],ymm8[4,5,6],ymm5[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, 32(%r8), %ymm4, %ymm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %ymm0
-; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %ymm4
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vmovaps %ymm4, %ymm6
-; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %ymm1
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm1[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm8
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm0[1],ymm4[2,3,4],ymm0[5],ymm4[6,7]
+; AVX2-FAST-NEXT:    vmovaps %ymm0, %ymm1
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[2,2,2,2]
+; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %ymm9
+; AVX2-FAST-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %ymm0
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vmovaps %ymm0, %ymm10
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm9[1,2],ymm5[3,4],ymm9[5,6],ymm5[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 48(%r8), %ymm5
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0],ymm4[1,2,3,4],ymm5[5],ymm4[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm8[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm1[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2],ymm5[3,4],ymm4[5,6],ymm5[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm10[2],ymm9[2],ymm10[3],ymm9[3],ymm10[6],ymm9[6],ymm10[7],ymm9[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[3,3,3,3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[2,1,3,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 56(%r8), %ymm5
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm5[2],ymm4[3,4,5,6],ymm5[7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
+; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
+; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm2, %ymm4
+; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm14, %ymm5
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6],ymm4[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, 64(%r8), %ymm3, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %ymm0
+; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %ymm1
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm1[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm5
+; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm0[1],ymm3[2,3,4],ymm0[5],ymm3[6,7]
-; AVX2-FAST-NEXT:    vmovaps %ymm0, %ymm5
+; AVX2-FAST-NEXT:    vmovaps %ymm0, %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[2,2,2,2]
-; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %ymm0
+; AVX2-FAST-NEXT:    vmovaps 64(%rdi), %ymm6
+; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 64(%rsi), %ymm0
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm0[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, %ymm7
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm1[1,2],ymm4[3,4],ymm1[5,6],ymm4[7]
-; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm6[1,2],ymm4[3,4],ymm6[5,6],ymm4[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 48(%r8), %ymm4
+; AVX2-FAST-NEXT:    vbroadcastsd 80(%r8), %ymm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0],ymm3[1,2,3,4],ymm4[5],ymm3[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm6[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm5[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm1[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2],ymm4[3,4],ymm3[5,6],ymm4[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm4 = ymm7[2],ymm1[2],ymm7[3],ymm1[3],ymm7[6],ymm1[6],ymm7[7],ymm1[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm4 = ymm7[2],ymm6[2],ymm7[3],ymm6[3],ymm7[6],ymm6[6],ymm7[7],ymm6[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[3,3,3,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[2,1,3,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 56(%r8), %ymm4
+; AVX2-FAST-NEXT:    vbroadcastsd 88(%r8), %ymm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm4[2],ymm3[3,4,5,6],ymm4[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
+; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
-; AVX2-FAST-NEXT:    # xmm3 = xmm1[0],mem[0],xmm1[1],mem[1]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
-; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm15, %ymm3
+; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm14, %ymm3
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5,6],ymm0[7]
-; AVX2-FAST-NEXT:    vmovaps 64(%rdi), %ymm4
-; AVX2-FAST-NEXT:    vinsertf128 $1, 64(%r8), %ymm4, %ymm3
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3],ymm3[4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, 96(%r8), %ymm1, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %ymm1
-; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %ymm3
+; AVX2-FAST-NEXT:    vmovaps 96(%rdx), %ymm1
+; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %ymm3
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vmovaps %ymm3, %ymm5
+; AVX2-FAST-NEXT:    vmovaps %ymm3, %ymm4
 ; 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 %ymm1, %ymm3
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-NEXT:    vmovaps 64(%rsi), %ymm6
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm6[1,1,2,2,5,5,6,6]
-; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2],ymm3[3,4],ymm4[5,6],ymm3[7]
-; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[2,2,2,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 80(%r8), %ymm3
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3,4],ymm3[5],ymm0[6,7]
-; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm1[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1,2],ymm3[3,4],ymm0[5,6],ymm3[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm3 = ymm6[2],ymm4[2],ymm6[3],ymm4[3],ymm6[6],ymm4[6],ymm6[7],ymm4[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 88(%r8), %ymm3
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2],ymm0[3,4,5,6],ymm3[7]
-; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
-; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
-; AVX2-FAST-NEXT:    # xmm2 = xmm1[0],mem[0],xmm1[1],mem[1]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
-; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm15, %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6],ymm0[7]
-; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %ymm1
-; AVX2-FAST-NEXT:    vinsertf128 $1, 96(%r8), %ymm1, %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3],ymm2[4],ymm0[5,6,7]
-; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 96(%rdx), %ymm13
-; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %ymm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vmovaps %ymm2, %ymm4
-; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1],ymm0[2,3,4],ymm13[5],ymm0[6,7]
-; AVX2-FAST-NEXT:    vmovaps %ymm13, %ymm3
-; AVX2-FAST-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-NEXT:    vmovaps 96(%rsi), %ymm5
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %ymm5
 ; AVX2-FAST-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
-; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm13
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,2,2,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 112(%r8), %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4],ymm2[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 96(%rsi), %ymm6
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1,2],ymm1[3,4],ymm5[5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 112(%r8), %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm4[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3,4],ymm0[5,6],ymm2[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm2 = ymm5[2],ymm1[2],ymm5[3],ymm1[3],ymm5[6],ymm1[6],ymm5[7],ymm1[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[3,3,3,3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm5[2],ymm6[3],ymm5[3],ymm6[6],ymm5[6],ymm6[7],ymm5[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 120(%r8), %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 120(%r8), %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
-; AVX2-FAST-NEXT:    # xmm2 = xmm1[0],mem[0],xmm1[1],mem[1]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
-; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm15, %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6],ymm0[7]
-; AVX2-FAST-NEXT:    vmovaps 128(%rdi), %ymm9
-; AVX2-FAST-NEXT:    vinsertf128 $1, 128(%r8), %ymm9, %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3],ymm2[4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm14, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, 128(%r8), %ymm1, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 128(%rdx), %ymm1
-; AVX2-FAST-NEXT:    vmovaps 128(%rcx), %ymm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vmovaps %ymm2, %ymm3
-; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-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 128(%rdx), %ymm13
+; AVX2-FAST-NEXT:    vmovaps 128(%rcx), %ymm1
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm3
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1],ymm0[2,3,4],ymm13[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps %ymm13, %ymm5
+; AVX2-FAST-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-NEXT:    vmovaps 128(%rsi), %ymm4
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vmovaps 128(%rdi), %ymm4
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm9[1,2],ymm2[3,4],ymm9[5,6],ymm2[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,2,2,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 144(%r8), %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4],ymm2[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 128(%rsi), %ymm13
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 144(%r8), %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3,4],ymm0[5,6],ymm2[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm2 = ymm4[2],ymm9[2],ymm4[3],ymm9[3],ymm4[6],ymm9[6],ymm4[7],ymm9[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[3,3,3,3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm13[2],ymm4[2],ymm13[3],ymm4[3],ymm13[6],ymm4[6],ymm13[7],ymm4[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 152(%r8), %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 152(%r8), %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-NEXT:    vunpcklps (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    vmovaps (%rsp), %xmm0 # 16-byte Reload
+; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
-; AVX2-FAST-NEXT:    # xmm2 = xmm1[0],mem[0],xmm1[1],mem[1]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
-; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm15, %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6],ymm0[7]
-; AVX2-FAST-NEXT:    vmovaps 160(%rdi), %ymm5
-; AVX2-FAST-NEXT:    vinsertf128 $1, 160(%r8), %ymm5, %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3],ymm2[4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm14, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, 160(%r8), %ymm1, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 160(%rdx), %ymm1
 ; AVX2-FAST-NEXT:    vmovaps 160(%rcx), %ymm12
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm12[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
-; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm3
+; AVX2-FAST-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-NEXT:    vmovaps 160(%rsi), %ymm10
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[1,1,2,2,5,5,6,6]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1,2],ymm2[3,4],ymm5[5,6],ymm2[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,2,2,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 176(%r8), %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4],ymm2[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 160(%rdi), %ymm10
+; AVX2-FAST-NEXT:    vmovaps 160(%rsi), %ymm9
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1,2],ymm1[3,4],ymm10[5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 176(%r8), %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm12[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3,4],ymm0[5,6],ymm2[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm2 = ymm10[2],ymm5[2],ymm10[3],ymm5[3],ymm10[6],ymm5[6],ymm10[7],ymm5[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[3,3,3,3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[6],ymm10[6],ymm9[7],ymm10[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 184(%r8), %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3,4,5,6],ymm2[7]
-; AVX2-FAST-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 184(%r8), %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
-; AVX2-FAST-NEXT:    # xmm2 = xmm1[0],mem[0],xmm1[1],mem[1]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
-; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm15, %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6],ymm0[7]
-; AVX2-FAST-NEXT:    vmovaps 192(%rdi), %ymm2
-; AVX2-FAST-NEXT:    vinsertf128 $1, 192(%r8), %ymm2, %ymm3
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3],ymm3[4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-FAST-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm14, %ymm3
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, 192(%r8), %ymm1, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 192(%rdx), %ymm8
-; AVX2-FAST-NEXT:    vmovaps 192(%rcx), %ymm7
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1],ymm0[2,3,4],ymm8[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 192(%rdx), %ymm7
+; AVX2-FAST-NEXT:    vmovaps 192(%rcx), %ymm6
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3,4],ymm7[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-NEXT:    vmovaps 192(%rsi), %ymm6
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm6[1,1,2,2,5,5,6,6]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm2[1,2],ymm3[3,4],ymm2[5,6],ymm3[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[2,2,2,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 208(%r8), %ymm3
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3,4],ymm3[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vmovaps 192(%rdi), %ymm5
+; AVX2-FAST-NEXT:    vmovaps 192(%rsi), %ymm4
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1,2],ymm1[3,4],ymm5[5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 208(%r8), %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm8[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1,2],ymm3[3,4],ymm0[5,6],ymm3[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm3 = ymm6[2],ymm2[2],ymm6[3],ymm2[3],ymm6[6],ymm2[6],ymm6[7],ymm2[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[3,3,3,3]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 216(%r8), %ymm3
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2],ymm0[3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 216(%r8), %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-FAST-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
-; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm15, %ymm1
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6],ymm0[7]
-; AVX2-FAST-NEXT:    vmovaps 224(%rdi), %ymm0
-; AVX2-FAST-NEXT:    vinsertf128 $1, 224(%r8), %ymm0, %ymm3
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0],ymm1[1,2,3],ymm3[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 224(%rdx), %ymm4
-; AVX2-FAST-NEXT:    vmovaps 224(%rcx), %ymm3
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm1[2,2,2,2]
-; AVX2-FAST-NEXT:    vmovaps 224(%rsi), %ymm1
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm1[1,1,2,2,5,5,6,6]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm0[1,2],ymm15[3,4],ymm0[5,6],ymm15[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0,1,2],ymm11[3,4],ymm15[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 240(%r8), %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0],ymm11[1,2,3,4],ymm15[5],ymm11[6,7]
-; AVX2-FAST-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0],ymm11[1,2],ymm15[3,4],ymm11[5,6],ymm15[7]
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,1,3,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm15[3,4],ymm11[5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 248(%r8), %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2],ymm11[3,4,5,6],ymm15[7]
-; AVX2-FAST-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm11 = mem[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm11 = ymm11[0,1,2,3],mem[4],ymm11[5,6,7]
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm14, %ymm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    vinsertf128 $1, 224(%r8), %ymm1, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vmovaps 224(%rdx), %ymm3
+; AVX2-FAST-NEXT:    vmovaps 224(%rcx), %ymm2
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4],ymm3[5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm0[2,2,2,2]
+; AVX2-FAST-NEXT:    vmovaps 224(%rdi), %ymm1
+; AVX2-FAST-NEXT:    vmovaps 224(%rsi), %ymm0
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm1[1,2],ymm11[3,4],ymm1[5,6],ymm11[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0,1,2],ymm8[3,4],ymm11[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 240(%r8), %ymm11
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0],ymm8[1,2,3,4],ymm11[5],ymm8[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0],ymm8[1,2],ymm11[3,4],ymm8[5,6],ymm11[7]
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm11 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[3,3,3,3]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,1,3,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm11[3,4],ymm8[5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 248(%r8), %ymm11
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm11[2],ymm8[3,4,5,6],ymm11[7]
+; AVX2-FAST-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = mem[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm15[4],ymm8[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm11 = mem[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = mem[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm14[0,1],ymm11[2],ymm14[3,4,5,6],ymm11[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm11[1,2,3],ymm8[4,5],ymm11[6,7]
+; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = ymm8[0,1,2],mem[3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = mem[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = ymm8[0,1,2,3],mem[4],ymm8[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm11 = mem[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = mem[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm14[0,1],ymm11[2],ymm14[3,4,5,6],ymm11[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm11[1,2,3],ymm8[4,5],ymm11[6,7]
+; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = ymm8[0,1,2],mem[3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = mem[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = ymm8[0,1,2,3],mem[4],ymm8[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm11 = mem[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0,1],ymm11[2],ymm15[3,4,5,6],ymm11[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm11[1,2,3],ymm8[4,5],ymm11[6,7]
+; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm11 = ymm8[0,1,2],mem[3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = mem[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = ymm8[0,1,2,3],mem[4],ymm8[5,6,7]
 ; AVX2-FAST-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm15 = mem[0,1,3,0,4,5,7,4]
 ; AVX2-FAST-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm14 = mem[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3,4,5,6],ymm15[7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm14[1,2,3],ymm11[4,5],ymm14[6,7]
-; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm11 = ymm11[0,1,2],mem[3],ymm11[4,5,6,7]
-; AVX2-FAST-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm11 = mem[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm11 = ymm11[0,1,2,3],mem[4],ymm11[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm14 = mem[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4,5,6],ymm14[7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm14[1,2,3],ymm11[4,5],ymm14[6,7]
-; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm11 = ymm11[0,1,2],mem[3],ymm11[4,5,6,7]
-; AVX2-FAST-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm11 = mem[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm11 = ymm11[0,1,2,3],mem[4],ymm11[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm14 = mem[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4,5,6],ymm14[7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm14[1,2,3],ymm11[4,5],ymm14[6,7]
-; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm11 = ymm11[0,1,2],mem[3],ymm11[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm14 = mem[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm13[4],ymm14[5,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm14[1,2,3],ymm8[4,5],ymm14[6,7]
+; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm8 = ymm8[0,1,2],mem[3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm14 = ymm13[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm14 = ymm14[0,1,2,3],mem[4],ymm14[5,6,7]
 ; AVX2-FAST-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm15 = mem[0,1,3,0,4,5,7,4]
 ; AVX2-FAST-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm13 = mem[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2],ymm13[3,4,5,6],ymm15[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0],ymm13[1,2,3],ymm14[4,5],ymm13[6,7]
-; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm14 = ymm13[0,1,2],mem[3],ymm13[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm13 = mem[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm13[0,1,2,3],ymm9[4],ymm13[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm13 = mem[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1],ymm13[2],ymm15[3,4,5,6],ymm13[7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm13[1,2,3],ymm9[4,5],ymm13[6,7]
-; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm9 = ymm9[0,1,2],mem[3],ymm9[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0,1,2,3],ymm5[4],ymm10[5,6,7]
+; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm13 = ymm13[0,1,2],mem[3],ymm13[4,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4],ymm9[5,6,7]
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm12[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    vpermilps $78, (%rsp), %ymm12 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm12 = mem[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2],ymm12[3,4,5,6],ymm10[7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm10[1,2,3],ymm5[4,5],ymm10[6,7]
-; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm5 = ymm5[0,1,2],mem[3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4],ymm6[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm8[2,3,0,1,6,7,4,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2],ymm7[3,4,5,6],ymm6[7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm6[1,2,3],ymm2[4,5],ymm6[6,7]
-; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,0,2,3,7,4,6,7]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[2,3,0,1,6,7,4,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2],ymm3[3,4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4,5],ymm10[6,7]
+; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm9 = ymm9[0,1,2],mem[3],ymm9[4,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm6[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2],ymm6[3,4,5,6],ymm5[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4,5],ymm5[6,7]
+; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT:    # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, 1184(%r9)
-; AVX2-FAST-NEXT:    vmovaps %ymm2, 1024(%r9)
-; AVX2-FAST-NEXT:    vmovaps %ymm5, 864(%r9)
-; AVX2-FAST-NEXT:    vmovaps %ymm9, 704(%r9)
-; AVX2-FAST-NEXT:    vmovaps %ymm14, 544(%r9)
+; AVX2-FAST-NEXT:    vmovaps %ymm4, 1024(%r9)
+; AVX2-FAST-NEXT:    vmovaps %ymm9, 864(%r9)
+; AVX2-FAST-NEXT:    vmovaps %ymm13, 704(%r9)
+; AVX2-FAST-NEXT:    vmovaps %ymm8, 544(%r9)
 ; AVX2-FAST-NEXT:    vmovaps %ymm11, 384(%r9)
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, 224(%r9)
@@ -5909,7 +5882,7 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, 1056(%r9)
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, 960(%r9)
-; AVX2-FAST-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, 928(%r9)
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, 896(%r9)
@@ -5967,87 +5940,83 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX2-FAST-PERLANE-LABEL: store_i32_stride5_vf64:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    subq $1768, %rsp # imm = 0x6E8
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %ymm14
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %ymm11
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT:    subq $1736, %rsp # imm = 0x6C8
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r8), %ymm15
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r8), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r8), %ymm14
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%r8), %ymm13
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%r8), %ymm12
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %xmm4
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %xmm7
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %xmm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rcx), %xmm0
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %xmm5
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %xmm8
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdx), %xmm1
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm7
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm8
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0,1],ymm6[2,3],ymm9[4,5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, (%r8), %ymm10, %ymm9
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0],ymm6[1,2,3],ymm9[4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm9
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %xmm6
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm7 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %xmm8
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %xmm2
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm10
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm12[0,1],ymm5[2,3],ymm12[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, (%r8), %ymm11, %ymm11
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm11[0],ymm5[1,2,3],ymm11[4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2],xmm7[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm15[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3,4,5],ymm8[6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,1,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3],ymm4[4,5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm15[0,1,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4,5],ymm5[6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm8[0],xmm6[0],xmm8[1],xmm6[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 32(%r8), %ymm11, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2,3],ymm5[4],ymm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %xmm4
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,1,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm5[2,3],ymm2[4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm13[0,1,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3,4,5],ymm3[6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 64(%r8), %ymm14, %ymm3
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm14, %ymm9
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3],ymm3[4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm6[2],xmm4[2],xmm6[3],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm12[0],xmm6[0],xmm12[1],xmm6[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm8[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 32(%r8), %ymm8, %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1,2,3],ymm8[4],ymm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm12[2],xmm6[2],xmm12[3],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3],ymm3[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm14[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4,5],ymm4[6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm4[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3],ymm6[4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 64(%r8), %ymm4, %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3],ymm4[4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm12[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm13[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rcx), %xmm0
@@ -6059,22 +6028,21 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 96(%r8), %ymm6, %ymm5
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%r8), %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 96(%r8), %ymm5, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%r8), %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rcx), %xmm0
@@ -6086,22 +6054,21 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rdi), %ymm7
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 128(%r8), %ymm7, %ymm5
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 128(%r8), %ymm5, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%r8), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%r8), %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm3[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rcx), %xmm0
@@ -6113,11 +6080,9 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 160(%r8), %ymm8, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, %ymm15
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 160(%r8), %ymm5, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
@@ -6141,11 +6106,9 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 192(%r8), %ymm8, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 192(%r8), %ymm5, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
@@ -6169,10 +6132,8 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 224(%r8), %ymm5, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6194,22 +6155,22 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm14[1],ymm0[2,3,4],ymm14[5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm14, %ymm3
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %ymm14
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm14[1,1,2,2,5,5,6,6]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1,2],ymm1[3,4],ymm10[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1,2],ymm1[3,4],ymm3[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 16(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm14[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm14[2],ymm10[2],ymm14[3],ymm10[3],ymm14[6],ymm10[6],ymm14[7],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[6],ymm3[6],ymm4[7],ymm3[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
@@ -6225,11 +6186,12 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1,2],ymm1[3,4],ymm11[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %ymm5
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 48(%r8), %ymm1
@@ -6238,7 +6200,7 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm11[2],ymm4[3],ymm11[3],ymm4[6],ymm11[6],ymm4[7],ymm11[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
@@ -6254,11 +6216,12 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1,2],ymm1[3,4],ymm9[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %ymm5
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 80(%r8), %ymm1
@@ -6267,7 +6230,7 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm9[2],ymm4[3],ymm9[3],ymm4[6],ymm9[6],ymm4[7],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
@@ -6283,10 +6246,12 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rsi), %ymm12
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[1,1,2,2,5,5,6,6]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2],ymm1[3,4],ymm6[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rsi), %ymm5
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 112(%r8), %ymm1
@@ -6295,7 +6260,7 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm12[2],ymm6[2],ymm12[3],ymm6[3],ymm12[6],ymm6[6],ymm12[7],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
@@ -6303,201 +6268,205 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rdx), %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rcx), %ymm11
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rcx), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-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 %ymm1, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rsi), %ymm10
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm10[1,1,2,2,5,5,6,6]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm7[1,2],ymm1[3,4],ymm7[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rsi), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 144(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm10[2],ymm7[2],ymm10[3],ymm7[3],ymm10[6],ymm7[6],ymm10[7],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm13[2],ymm4[2],ymm13[3],ymm4[3],ymm13[6],ymm4[6],ymm13[7],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 152(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rdx), %ymm9
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rcx), %ymm8
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm9[1],ymm0[2,3,4],ymm9[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rdx), %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rcx), %ymm11
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rsi), %ymm7
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[1,1,2,2,5,5,6,6]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1,2],ymm1[3,4],ymm15[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rdi), %ymm9
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rsi), %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1,2],ymm1[3,4],ymm9[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 176(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm8[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm7[2],ymm15[2],ymm7[3],ymm15[3],ymm7[6],ymm15[6],ymm7[7],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 184(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdx), %ymm6
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rcx), %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm6[1],ymm0[2,3,4],ymm6[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdx), %ymm7
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rcx), %ymm6
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3,4],ymm7[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdi), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rsi), %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm13, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1,2],ymm1[3,4],ymm13[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1,2],ymm1[3,4],ymm5[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 208(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm13[2],ymm4[3],ymm13[3],ymm4[6],ymm13[6],ymm4[7],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 216(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdx), %ymm13
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rcx), %ymm3
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1],ymm0[2,3,4],ymm13[5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rsi), %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm2[1,1,2,2,5,5,6,6]
-; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm0[1,2],ymm15[3,4],ymm0[5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdx), %ymm3
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rcx), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4],ymm3[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm0[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rsi), %ymm0
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm1[1,2],ymm15[3,4],ymm1[5,6],ymm15[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0,1,2],ymm1[3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0,1,2],ymm10[3,4],ymm15[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 240(%r8), %ymm15
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0],ymm1[1,2,3,4],ymm15[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm13[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0],ymm1[1,2],ymm15[3,4],ymm1[5,6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[6],ymm0[6],ymm2[7],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0],ymm10[1,2,3,4],ymm15[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm10, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0],ymm10[1,2],ymm15[3,4],ymm10[5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm1[2,1,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,1,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm15[3,4],ymm10[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 248(%r8), %ymm15
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm15[2],ymm0[3,4,5,6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm14[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm15[2],ymm10[3,4,5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = mem[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = ymm10[0,1,2,3],mem[4],ymm10[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3,4,5,6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm14[1,2,3],ymm0[4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm14[1,2,3],ymm10[4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = ymm10[0,1,2],mem[3],ymm10[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = mem[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = ymm10[0,1,2,3],mem[4],ymm10[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[0,1,3,0,4,5,7,4]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4,5,6],ymm14[7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm14[1,2,3],ymm0[4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm14[1,2,3],ymm10[4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = ymm10[0,1,2],mem[3],ymm10[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = mem[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = ymm10[0,1,2,3],mem[4],ymm10[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[0,1,3,0,4,5,7,4]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4,5,6],ymm14[7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm14[1,2,3],ymm0[4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm15 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm12[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm12 = mem[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[2,3,0,1,6,7,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm14[0,1],ymm12[2],ymm14[3,4,5,6],ymm12[7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1,2,3],ymm0[4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm12 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm10[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm11[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm11 = mem[2,3,0,1,6,7,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2],ymm11[3,4,5,6],ymm10[7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm10[1,2,3],ymm0[4,5],ymm10[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm7 = ymm7[0,1,2,3],mem[4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,3,0,1,6,7,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2],ymm9[3,4,5,6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1,2,3],ymm7[4,5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm14[1,2,3],ymm10[4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = ymm10[0,1,2],mem[3],ymm10[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $227, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = mem[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = ymm10[0,1,2,3],mem[4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4,5,6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm14[1,2,3],ymm10[4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = ymm10[0,1,2],mem[3],ymm10[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm13 = ymm13[0,1,2,3],mem[4],ymm13[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4,5,6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1,2,3],ymm13[4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm13 = ymm13[0,1,2],mem[3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm11[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1],ymm9[2],ymm11[3,4,5,6],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4,5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm8 = ymm8[0,1,2],mem[3],ymm8[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $16, (%rsp), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm4 = ymm4[0,1,2,3],mem[4],ymm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm6[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2],ymm6[3,4,5,6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4,5],ymm5[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[3,0,2,3,7,4,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm13[2,3,0,1,6,7,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm1[0,1,2],mem[3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, 1184(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 1184(%r9)
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm4, 1024(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm7, 864(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 704(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm12, 544(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm15, 384(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, 864(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm13, 704(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm10, 544(%r9)
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 384(%r9)
 ; 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, 64(%r9)
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 1248(%r9)
-; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 1216(%r9)
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 1088(%r9)
@@ -6559,7 +6528,7 @@ define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; 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:    addq $1768, %rsp # imm = 0x6E8
+; AVX2-FAST-PERLANE-NEXT:    addq $1736, %rsp # imm = 0x6C8
 ; AVX2-FAST-PERLANE-NEXT:    vzeroupper
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;

diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
index e0de264..8f2eeed 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
@@ -7212,29 +7212,29 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride7_vf64:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $3384, %rsp # imm = 0xD38
+; AVX1-ONLY-NEXT:    subq $3432, %rsp # imm = 0xD68
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm5
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %ymm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %ymm0
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%r8), %ymm4
+; AVX1-ONLY-NEXT:    vmovaps 224(%r8), %ymm3
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rax), %ymm4
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rax), %ymm5
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,0],ymm3[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm5[1],ymm2[3],ymm5[3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,0],ymm5[4,5],ymm1[6,4]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm2
@@ -7288,8 +7288,8 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %ymm13
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm13[0],ymm2[0],ymm13[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovaps (%r9), %ymm12
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm12[0],ymm2[0],ymm12[2],ymm2[2]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,1],ymm1[6,4],ymm2[6,5]
 ; AVX1-ONLY-NEXT:    vmovaps (%rax), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7299,8 +7299,8 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm6
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm6[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm4
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm4[0],xmm1[0]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, %xmm7
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7310,15 +7310,15 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, %xmm8
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm10
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7326,14 +7326,14 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm5[1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm6[1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm6[1]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm6[1,1],xmm1[0,2]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[1,1],xmm1[0,2]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm10[1],xmm8[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm9[1],xmm8[1],zero
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7343,11 +7343,11 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm1[1,1],ymm0[5,5],ymm1[5,5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %ymm7
-; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %ymm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[1,1],ymm1[1,1],ymm7[5,5],ymm1[5,5]
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %ymm8
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm8[1,1],ymm1[5,5],ymm8[5,5]
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7366,23 +7366,23 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm6
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm6[0],xmm1[0]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,1]
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, %xmm5
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm2
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, %xmm8
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, %xmm7
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7390,14 +7390,14 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm5[1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm4[1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm6[1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm6[1,1],xmm1[0,2]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm9[1],xmm8[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm9[1],xmm7[1],zero
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7415,8 +7415,8 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %ymm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm14[0],ymm2[0],ymm14[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %ymm15
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm15[0],ymm2[0],ymm15[2],ymm2[2]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,1],ymm1[6,4],ymm2[6,5]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7426,8 +7426,8 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm4
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm4[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm5
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm5[0],xmm1[0]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, %xmm6
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7435,7 +7435,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm2
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, %xmm8
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, %xmm7
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7443,9 +7443,9 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rax), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7453,14 +7453,14 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm5[1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm4[1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm6[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[1,1],xmm1[0,2]
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm6[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm5[1,1],xmm1[0,2]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm9[1],xmm8[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm9[1],xmm7[1],zero
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7470,17 +7470,16 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,1],ymm0[1,1],ymm1[5,5],ymm0[5,5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %ymm2
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %ymm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %ymm15
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm15[1,1],ymm1[5,5],ymm15[5,5]
-; AVX1-ONLY-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm1[1,1],ymm2[5,5],ymm1[5,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %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:    vmovaps 96(%r9), %ymm10
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm10[0],ymm2[0],ymm10[2],ymm2[2]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,1],ymm1[6,4],ymm2[6,5]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rax), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7501,15 +7500,15 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, %xmm6
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm8
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %xmm9
-; AVX1-ONLY-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm9[0],xmm3[0],xmm9[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rax), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7517,14 +7516,14 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm9[1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm7[1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm4[1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[1,1],xmm1[0,2]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm8[1],xmm6[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm9[1],xmm6[1],zero
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7565,15 +7564,15 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, %xmm6
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm8
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 160(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %xmm9
-; AVX1-ONLY-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm9[0],xmm3[0],xmm9[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rax), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7581,35 +7580,34 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm9[1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm7[1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm4[1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[1,1],xmm1[0,2]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm8[1],xmm6[1],zero
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm9[1],xmm6[1],zero
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[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 160(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,1],ymm0[1,1],ymm1[5,5],ymm0[5,5]
+; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm1[1,1],ymm0[5,5],ymm1[5,5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %ymm10
-; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %ymm9
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[1,1],ymm9[1,1],ymm10[5,5],ymm9[5,5]
-; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
-; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %ymm12
-; AVX1-ONLY-NEXT:    vmovaps 160(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %ymm6
+; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %ymm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm12[0],ymm1[2],ymm12[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm12[2,1],ymm1[6,4],ymm12[6,5]
-; AVX1-ONLY-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[1,1],ymm1[1,1],ymm6[5,5],ymm1[5,5]
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
+; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %ymm9
+; AVX1-ONLY-NEXT:    vmovaps 160(%r9), %ymm13
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm13[0],ymm9[0],ymm13[2],ymm9[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm9[2,1],ymm1[6,4],ymm9[6,5]
+; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rax), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm1[2,3]
@@ -7617,11 +7615,12 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm11
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm11[0],xmm6[0]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm6[2,1]
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm11[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,1]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, %xmm14
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm5
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm4
@@ -7632,9 +7631,9 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rax), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7642,25 +7641,24 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm8[1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm7[1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm6[1],xmm11[1]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm14[1],xmm11[1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm11[1,1],xmm1[0,2]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm4[1],xmm5[1],zero
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm4
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %ymm5
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm5[1,1],ymm8[1,1],ymm5[5,5],ymm8[5,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm5[1,1],ymm4[1,1],ymm5[5,5],ymm4[5,5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %ymm6
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm6[1,1],ymm1[5,5],ymm6[5,5]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %ymm7
+; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %ymm14
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[1,1],ymm14[1,1],ymm7[5,5],ymm14[5,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %ymm11
 ; AVX1-ONLY-NEXT:    vmovaps 192(%r9), %ymm3
@@ -7671,18 +7669,18 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm4[1],ymm0[3],ymm4[3]
+; 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:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %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:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm13[1],ymm1[3],ymm13[3]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[1,1],ymm1[0,2],ymm13[5,5],ymm1[4,6]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm12[1],ymm1[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[1,1],ymm1[0,2],ymm12[5,5],ymm1[4,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rax), %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7]
@@ -7692,8 +7690,8 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; 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:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm1 = ymm7[2],mem[2],ymm7[3],mem[3],ymm7[6],mem[6],ymm7[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm8[2],ymm1[3],ymm8[3],ymm1[6],ymm8[6],ymm1[7],ymm8[7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[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
@@ -7708,30 +7706,31 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; 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:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm1 = ymm7[2],mem[2],ymm7[3],mem[3],ymm7[6],mem[6],ymm7[7],mem[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm8[2],mem[2],ymm8[3],mem[3],ymm8[6],mem[6],ymm8[7],mem[7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm14[1],ymm1[3],ymm14[3]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm14[1,1],ymm1[0,2],ymm14[5,5],ymm1[4,6]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm15[1],ymm1[3],ymm15[3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[1,1],ymm1[0,2],ymm15[5,5],ymm1[4,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rax), %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[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:    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), %ymm15 # 32-byte Reload
+; 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:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm14[2],ymm15[2],ymm14[3],ymm15[3],ymm14[6],ymm15[6],ymm14[7],ymm15[7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; 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:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm1[0,2],ymm2[5,5],ymm1[4,6]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm10[1],ymm1[3],ymm10[3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[1,1],ymm1[0,2],ymm10[5,5],ymm1[4,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 112(%rax), %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7]
@@ -7758,24 +7757,26 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; 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:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm10[2],ymm9[2],ymm10[3],ymm9[3],ymm10[6],ymm9[6],ymm10[7],ymm9[7]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm12[1],ymm2[1],ymm12[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm1[0,2],ymm2[5,5],ymm1[4,6]
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm2[2],ymm6[3],ymm2[3],ymm6[6],ymm2[6],ymm6[7],ymm2[7]
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, %ymm10
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm9[1],ymm13[1],ymm9[3],ymm13[3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[1,1],ymm1[0,2],ymm13[5,5],ymm1[4,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 176(%rax), %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps %ymm5, %ymm9
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, %ymm6
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm8[1],ymm5[3],ymm8[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm6[2],ymm5[3],ymm6[3],ymm5[6],ymm6[6],ymm5[7],ymm6[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm7[2],ymm14[2],ymm7[3],ymm14[3],ymm7[6],ymm14[6],ymm7[7],ymm14[7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm11[1],ymm3[1],ymm11[3],ymm3[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[1,1],ymm1[0,2],ymm3[5,5],ymm1[4,6]
@@ -7784,49 +7785,49 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm11
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm11[1],xmm1[1]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,1],xmm0[0,2]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm11[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[1,1],xmm0[0,2]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm13 = zero,xmm0[1],xmm2[1],zero
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm0 = zero,xmm1[1],xmm4[1],zero
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm0[1,2],ymm3[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 228(%r8), %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 228(%r9), %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm13[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rax), %ymm0, %ymm13
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm13[1],ymm3[2,3,4],ymm13[5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm2[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,0],xmm0[2,3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm11[3,3],xmm1[3,3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm13[5,6],ymm3[7]
-; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm3 = ymm3[0,1],mem[2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastss 232(%r9), %xmm13
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastss 232(%rax), %ymm13
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm13[4],ymm3[5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rax), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3,4],ymm0[5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm11[3,3],xmm2[3,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6],ymm0[7]
+; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 232(%r9), %xmm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 232(%rax), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm14[3,3],ymm7[3,3],ymm14[7,7],ymm7[7,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm6[3,3],ymm5[3,3],ymm6[7,7],ymm5[7,7]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm8[3,3],ymm9[3,3],ymm8[7,7],ymm9[7,7]
-; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm13[0,1,2],ymm3[3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastss 220(%r8), %ymm13
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm13[5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vbroadcastss 220(%r9), %ymm13
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 216(%rax), %ymm13
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm13[0],ymm3[1,2,3,4,5,6],ymm13[7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm11[0]
+; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 220(%r8), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastss 220(%r9), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 216(%rax), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3,4,5,6],ymm3[7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm11[0]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm11[2,1]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm2
@@ -7839,12 +7840,12 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[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:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm5[3,3],ymm3[3,3],ymm5[7,7],ymm3[7,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm4[3,3],ymm3[3,3],ymm4[7,7],ymm3[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[3,3],ymm2[3,3],ymm15[7,7],ymm2[7,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[3,3],ymm2[3,3],ymm6[7,7],ymm2[7,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 252(%r8), %ymm1
@@ -7854,9 +7855,9 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vbroadcastsd 248(%rax), %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm5[0],ymm3[2],ymm5[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm5[3,1],ymm0[0,2],ymm5[7,5],ymm0[4,6]
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm15[0],ymm2[0],ymm15[1],ymm2[1],ymm15[4],ymm2[4],ymm15[5],ymm2[5]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm4[3,1],ymm0[0,2],ymm4[7,5],ymm0[4,6]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm6[0],ymm2[0],ymm6[1],ymm2[1],ymm6[4],ymm2[4],ymm6[5],ymm2[5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 236(%r8), %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7]
@@ -7865,8 +7866,8 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[1,1],ymm5[1,1],ymm3[5,5],ymm5[5,5]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm15[1,1],ymm2[5,5],ymm15[5,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[1,1],ymm4[1,1],ymm3[5,5],ymm4[5,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm6[1,1],ymm2[5,5],ymm6[5,5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 240(%r8), %ymm1
@@ -7897,8 +7898,9 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[3,3],mem[3,3],ymm0[7,7],mem[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[3,3],ymm10[3,3],ymm4[7,7],ymm10[7,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,3],ymm7[3,3],ymm1[7,7],ymm7[7,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -7919,7 +7921,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
-; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vpermilps $170, (%rsp), %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]
@@ -7927,9 +7929,9 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[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), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm8[3,3],mem[3,3],ymm8[7,7],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[3,3],mem[3,3],ymm0[7,7],mem[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
@@ -7954,7 +7956,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
-; AVX1-ONLY-NEXT:    vpermilps $170, (%rsp), %xmm1 # 16-byte Folded Reload
+; 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]
@@ -7963,7 +7965,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[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[3,3],ymm7[3,3],ymm0[7,7],ymm7[7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,3],ymm8[3,3],ymm0[7,7],ymm8[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -7997,11 +7999,11 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[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[3,3],ymm14[3,3],ymm0[7,7],ymm14[7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,3],ymm12[3,3],ymm0[7,7],ymm12[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm15, %ymm4
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,3],ymm4[3,3],ymm1[7,7],ymm4[7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,3],ymm15[3,3],ymm1[7,7],ymm15[7,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -8030,9 +8032,9 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[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 $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[3,3],mem[3,3],ymm0[7,7],mem[7,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm14[3,3],mem[3,3],ymm14[7,7],mem[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -8057,21 +8059,20 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6],ymm2[7]
-; 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:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm15[0,1,2],xmm2[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm15[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm13[0,1,2],xmm2[3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 168(%rax), %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[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), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[3,3],ymm14[3,3],ymm13[7,7],ymm14[7,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm10[3,3],ymm11[3,3],ymm10[7,7],ymm11[7,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[3,3],ymm7[3,3],ymm9[7,7],ymm7[7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[3,3],ymm10[3,3],ymm9[7,7],ymm10[7,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -8092,8 +8093,8 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6],ymm2[7]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm11[2,2,2,2]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm8[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rax), %ymm3
@@ -8101,98 +8102,100 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[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:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm10[0],ymm0[1],ymm10[1],ymm0[4],ymm10[4],ymm0[5],ymm10[5]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm7[0],ymm0[1],ymm7[1],ymm0[4],ymm7[4],ymm0[5],ymm7[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm10[0],ymm2[2],ymm10[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm10[3,1],ymm2[0,2],ymm10[7,5],ymm2[4,6]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm7[0],ymm2[2],ymm7[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm7[3,1],ymm3[0,2],ymm7[7,5],ymm3[4,6]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm2 = xmm2[3,3],mem[3,3]
-; 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:    vblendps {{.*#+}} ymm3 = ymm0[0],ymm2[1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm2[3,3],mem[3,3]
+; 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:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[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), %ymm7 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm8[0],ymm2[2],ymm8[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm8[3,1],ymm2[0,2],ymm8[7,5],ymm2[4,6]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm2 = xmm2[3,3],mem[3,3]
-; 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:    vblendps {{.*#+}} ymm8 = ymm0[0],ymm2[1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm7[0],ymm2[2],ymm7[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm7[3,1],ymm3[0,2],ymm7[7,5],ymm3[4,6]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $255, (%rsp), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[3,3],mem[3,3]
+; 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:    vblendps {{.*#+}} ymm2 = ymm0[0],ymm3[1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[1],ymm6[1],ymm0[4],ymm6[4],ymm0[5],ymm6[5]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm6[0],ymm2[2],ymm6[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm6[3,1],ymm2[0,2],ymm6[7,5],ymm2[4,6]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vshufps $255, (%rsp), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm2 = xmm2[3,3],mem[3,3]
-; 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:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[1],ymm4[1],ymm2[4],ymm4[4],ymm2[5],ymm4[5]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm6[0],ymm3[2],ymm6[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm6[3,1],ymm3[0,2],ymm6[7,5],ymm3[4,6]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[3,3],mem[3,3]
+; 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:    vblendps {{.*#+}} ymm7 = ymm0[0],ymm3[1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 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), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm6[0],ymm4[0],ymm6[2],ymm4[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm4[3,1],ymm10[0,2],ymm4[7,5],ymm10[4,6]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3],ymm2[4,5],ymm10[6,7]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm10 = xmm4[3,3],mem[3,3]
-; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm10 = xmm10[0,1,2],mem[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm10[1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm10 = ymm12[0],ymm1[0],ymm12[1],ymm1[1],ymm12[4],ymm1[4],ymm12[5],ymm1[5]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[3,1],ymm3[0,2],ymm4[7,5],ymm3[4,6]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[3,3],mem[3,3]
+; 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:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm3 = ymm12[0],ymm1[0],ymm12[1],ymm1[1],ymm12[4],ymm1[4],ymm12[5],ymm1[5]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm4[0],ymm1[0],ymm4[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm1[3,1],ymm12[0,2],ymm1[7,5],ymm12[4,6]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm4[0],ymm14[0],ymm4[2],ymm14[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm14[3,1],ymm12[0,2],ymm14[7,5],ymm12[4,6]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm12[0,1,2,3],ymm3[4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm12 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm12 = xmm1[3,3],mem[3,3]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm12 = xmm12[0,1,2],mem[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm12[1,2,3],ymm10[4,5,6,7]
-; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm12 = ymm9[0],ymm7[0],ymm9[1],ymm7[1],ymm9[4],ymm7[4],ymm9[5],ymm7[5]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm14[0],ymm13[0],ymm14[2],ymm13[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm13[3,1],ymm14[0,2],ymm13[7,5],ymm14[4,6]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm12[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm12 = xmm15[3,3],mem[3,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm12[1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm12 = 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), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm11[0],ymm1[0],ymm11[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[3,1],ymm14[0,2],ymm1[7,5],ymm14[4,6]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm12[4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm13[3,3],xmm15[3,3]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm12 = xmm12[0,1,2],mem[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm12[1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm12[1,2,3],ymm10[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm12 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[4],mem[4],ymm1[5],mem[5]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm4[0],ymm1[0],ymm4[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm1[3,1],ymm14[0,2],ymm1[7,5],ymm14[4,6]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm12[4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm5[3,3],xmm11[3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm1[3,1],ymm14[0,2],ymm1[7,5],ymm14[4,6]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm12[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm5[3,3],xmm8[3,3]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm12 = xmm12[0,1,2],mem[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm12[1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm12[1,2,3],ymm9[4,5,6,7]
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps %ymm7, 1440(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm9, 1216(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm10, 992(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm2, 768(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm8, 320(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm9, 1440(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm10, 1216(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 992(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 768(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 544(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 320(%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, 1504(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -8291,37 +8294,37 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1632(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1600(%rax)
-; AVX1-ONLY-NEXT:    addq $3384, %rsp # imm = 0xD38
+; AVX1-ONLY-NEXT:    addq $3432, %rsp # imm = 0xD68
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: store_i32_stride7_vf64:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    subq $2984, %rsp # imm = 0xBA8
+; AVX2-SLOW-NEXT:    subq $2968, %rsp # imm = 0xB98
 ; AVX2-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX2-SLOW-NEXT:    vmovaps (%rax), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vmovaps (%r8), %xmm2
-; AVX2-SLOW-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 32(%r8), %xmm4
-; AVX2-SLOW-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps (%r8), %xmm5
+; AVX2-SLOW-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 32(%r8), %xmm3
+; AVX2-SLOW-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps (%r9), %xmm1
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 32(%r9), %xmm7
-; AVX2-SLOW-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 32(%r9), %xmm4
+; AVX2-SLOW-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps (%rcx), %xmm9
-; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %xmm3
-; AVX2-SLOW-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %xmm10
+; AVX2-SLOW-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps (%rdx), %xmm8
 ; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm8[1],xmm9[1],zero
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm6
-; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %xmm10
-; AVX2-SLOW-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %xmm7
+; AVX2-SLOW-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm5
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %xmm11
 ; AVX2-SLOW-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8334,15 +8337,15 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rax), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[1,1,1,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm11[1,1,2,2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm10[2],xmm1[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %xmm7
-; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm7[1],xmm3[1],zero
+; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm7[1],xmm10[1],zero
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8521,11 +8524,11 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %ymm1
-; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %ymm0
+; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %ymm0
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
+; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %ymm1
+; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rdx), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8661,13 +8664,13 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm14 = zero,xmm1[1],xmm0[1],zero
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1,2],ymm13[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vbroadcastss 228(%r8), %ymm14
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3],ymm13[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm14 = xmm3[1,1,1,1]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm14[4,5],ymm13[6,7]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm15, %ymm1, %ymm14
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 228(%r8), %ymm12
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm3[1,1,1,1]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5],ymm12[6,7]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm13
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3],xmm4[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
@@ -8683,37 +8686,37 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vbroadcastss 232(%rax), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps 224(%rdi), %ymm10
-; AVX2-SLOW-NEXT:    vmovaps 224(%rsi), %ymm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm4[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm10[1],ymm0[2,3,4],ymm10[5],ymm0[6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovaps 224(%rdx), %ymm12
-; AVX2-SLOW-NEXT:    vmovaps 224(%rcx), %ymm2
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm14 = ymm12[1,1],ymm2[1,1],ymm12[5,5],ymm2[5,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6],ymm13[7]
-; AVX2-SLOW-NEXT:    vbroadcastsd 240(%r8), %ymm14
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0],ymm13[1,2,3,4,5,6],ymm14[7]
-; AVX2-SLOW-NEXT:    vbroadcastss 240(%r9), %xmm14
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vbroadcastss 240(%rax), %ymm14
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm14[2],ymm13[3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps 224(%rdi), %ymm4
+; AVX2-SLOW-NEXT:    vmovaps 224(%rsi), %ymm2
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm0[2,2,2,2]
+; AVX2-SLOW-NEXT:    vmovaps 224(%rdx), %ymm10
+; AVX2-SLOW-NEXT:    vmovaps 224(%rcx), %ymm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm10[1,1],ymm1[1,1],ymm10[5,5],ymm1[5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm13[5,6],ymm12[7]
+; AVX2-SLOW-NEXT:    vbroadcastsd 240(%r8), %ymm13
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0],ymm12[1,2,3,4,5,6],ymm13[7]
+; AVX2-SLOW-NEXT:    vbroadcastss 240(%r9), %xmm13
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1],ymm12[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 240(%rax), %ymm13
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1],ymm13[2],ymm12[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vbroadcastss %xmm9, %xmm13
-; AVX2-SLOW-NEXT:    vbroadcastss %xmm8, %xmm14
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[0,1,2,2]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[0,1,2,1]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1],ymm13[2,3],ymm14[4,5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss %xmm9, %xmm12
+; AVX2-SLOW-NEXT:    vbroadcastss %xmm8, %xmm13
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX2-SLOW-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm14[0],ymm15[0],ymm14[2],ymm15[2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm14[4,5,6],ymm13[7]
-; AVX2-SLOW-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX2-SLOW-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm13[0],ymm14[0],ymm13[2],ymm14[2]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6],ymm12[7]
+; AVX2-SLOW-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[3,3],xmm5[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
@@ -8721,36 +8724,36 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 8(%rax), %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vbroadcastss %xmm1, %xmm5
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vbroadcastss %xmm12, %xmm5
 ; AVX2-SLOW-NEXT:    vbroadcastss %xmm7, %xmm6
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm13[0],xmm9[1],xmm13[1]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm15[0],xmm9[1],xmm15[1]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm8[0],ymm6[2],ymm8[2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm14[3,3],xmm15[3,3]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm7[2],xmm1[2],xmm7[3],xmm1[3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm13[3,3],xmm14[3,3]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm7[2],xmm12[2],xmm7[3],xmm12[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm13[2,2,2,2]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm15[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 40(%rax), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -8760,28 +8763,28 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vbroadcastss %xmm13, %xmm5
 ; AVX2-SLOW-NEXT:    vbroadcastss %xmm11, %xmm6
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm12[3,3],xmm9[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm11[2],xmm13[2],xmm11[3],xmm13[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 72(%rax), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
@@ -8791,9 +8794,9 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vbroadcastss %xmm14, %xmm6
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
@@ -8805,7 +8808,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm1[3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm12[3,3],xmm9[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
@@ -8817,41 +8820,41 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vbroadcastss %xmm12, %xmm5
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vbroadcastss %xmm13, %xmm5
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vbroadcastss %xmm14, %xmm6
+; AVX2-SLOW-NEXT:    vbroadcastss %xmm13, %xmm6
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm11[0],xmm9[0],xmm11[1],xmm9[1]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps (%rsp), %xmm0 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
-; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm11[3,3],xmm9[3,3]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm8[2,2,2,2]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm11[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 136(%rax), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vbroadcastss %xmm12, %xmm5
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vbroadcastss %xmm13, %xmm5
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vbroadcastss %xmm14, %xmm6
+; AVX2-SLOW-NEXT:    vbroadcastss %xmm13, %xmm6
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
@@ -8865,10 +8868,10 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
-; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
@@ -8884,39 +8887,39 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vbroadcastss %xmm13, %xmm6
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
+; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm8[3,3],xmm1[3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm11[2],xmm13[3],xmm11[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm9[2,2,2,2]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm12[0,1,2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 200(%rax), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[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:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm5 = ymm13[2],mem[2],ymm13[3],mem[3],ymm13[6],mem[6],ymm13[7],mem[7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm5 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[3,3,3,3]
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm6 = ymm9[2],ymm14[2],ymm9[3],ymm14[3],ymm9[6],ymm14[6],ymm9[7],ymm14[7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm6 = ymm13[2],ymm11[2],ymm13[3],ymm11[3],ymm13[6],ymm11[6],ymm13[7],ymm11[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 220(%r8), %ymm6
@@ -8927,17 +8930,17 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0],ymm5[1,2,3,4,5,6],ymm6[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vbroadcastss 240(%rdx), %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm2[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm1[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6],ymm6[7]
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm6 = ymm10[0],ymm4[0],ymm10[1],ymm4[1],ymm10[4],ymm4[4],ymm10[5],ymm4[5]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm6 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 236(%r8), %ymm6
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1],ymm5[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,2,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm10[2],ymm4[2],ymm10[3],ymm4[3],ymm10[6],ymm4[6],ymm10[7],ymm4[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[6],ymm2[6],ymm4[7],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm6 = ymm12[2],ymm2[2],ymm12[3],ymm2[3],ymm12[6],ymm2[6],ymm12[7],ymm2[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm6 = ymm10[2],ymm1[2],ymm10[3],ymm1[3],ymm10[6],ymm1[6],ymm10[7],ymm1[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
@@ -8950,8 +8953,8 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm7[2,3],ymm6[2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0],ymm3[1],ymm5[2,3,4],ymm3[5],ymm5[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm2[2],ymm12[2],ymm2[3],ymm12[3],ymm2[6],ymm12[6],ymm2[7],ymm12[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm10[2],ymm4[3],ymm10[3],ymm4[6],ymm10[6],ymm4[7],ymm10[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm10[2],ymm1[3],ymm10[3],ymm1[6],ymm10[6],ymm1[7],ymm10[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm4[2],ymm2[3],ymm4[3],ymm2[6],ymm4[6],ymm2[7],ymm4[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -8967,28 +8970,28 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[1,1],ymm3[1,1],ymm7[5,5],ymm3[5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,1],ymm3[1,1],ymm8[5,5],ymm3[5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[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), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm14[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[0,1,0,1,4,5,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3,4],ymm1[5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 16(%rax), %ymm2
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm15[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm10[1],ymm0[2,3,4],ymm10[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm6[1,1],ymm1[5,5],ymm6[5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm7[1,1],ymm1[5,5],ymm7[5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[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]
@@ -9000,14 +9003,14 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[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:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm15[1,1],ymm1[5,5],ymm15[5,5]
+; AVX2-SLOW-NEXT:    vshufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm1 = ymm1[1,1],mem[1,1],ymm1[5,5],mem[5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[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]
@@ -9077,9 +9080,9 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[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:    vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm13[1,1],ymm0[5,5],ymm13[5,5]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm12[1,1],ymm0[5,5],ymm12[5,5]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
@@ -9104,31 +9107,30 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    # xmm1 = xmm1[0,1,2],mem[3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm3[2],ymm7[2],ymm3[3],ymm7[3],ymm3[6],ymm7[6],ymm3[7],ymm7[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm3[2],ymm8[2],ymm3[3],ymm8[3],ymm3[6],ymm8[6],ymm3[7],ymm8[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[3,3],ymm11[3,3],ymm8[7,7],ymm11[7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[3,3],ymm14[3,3],ymm9[7,7],ymm14[7,7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 48(%rdx), %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
-; AVX2-SLOW-NEXT:    vmovaps %ymm10, %ymm2
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm10[0],ymm12[0],ymm10[1],ymm12[1],ymm10[4],ymm12[4],ymm10[5],ymm12[5]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm10[0],ymm15[0],ymm10[1],ymm15[1],ymm10[4],ymm15[4],ymm10[5],ymm15[5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = ymm6[2],mem[2],ymm6[3],mem[3],ymm6[6],mem[6],ymm6[7],mem[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm12[2],ymm2[2],ymm12[3],ymm2[3],ymm12[6],ymm2[6],ymm12[7],ymm2[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = ymm7[2],mem[2],ymm7[3],mem[3],ymm7[6],mem[6],ymm7[7],mem[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm15[2],ymm10[2],ymm15[3],ymm10[3],ymm15[6],ymm10[6],ymm15[7],ymm10[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -9138,24 +9140,23 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm15 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 80(%rdx), %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm15[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm9[0],ymm2[1],ymm9[1],ymm2[4],ymm9[4],ymm2[5],ymm9[5]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm6[0],ymm2[1],ymm6[1],ymm2[4],ymm6[4],ymm2[5],ymm6[5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = ymm15[2],mem[2],ymm15[3],mem[3],ymm15[6],mem[6],ymm15[7],mem[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm2[2],ymm9[3],ymm2[3],ymm9[6],ymm2[6],ymm9[7],ymm2[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm2[2],ymm6[3],ymm2[3],ymm6[6],ymm2[6],ymm6[7],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -9165,23 +9166,23 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT:    vbroadcastss 112(%rdx), %ymm0
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[3,1,2,0,7,5,6,4]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 112(%rdx), %ymm2
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm1[4,5],ymm6[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm0 = ymm9[2],mem[2],ymm9[3],mem[3],ymm9[6],mem[6],ymm9[7],mem[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm1[1,2,3],ymm6[4,5,6,7]
+; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -9189,92 +9190,90 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm9[0],ymm1[1,2],ymm9[3,4],ymm1[5,6],ymm9[7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3,4],ymm1[5,6],ymm8[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 144(%rdx), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm0[3,1,2,0,7,5,6,4]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5],ymm1[6],ymm9[7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3,4,5],ymm1[6],ymm8[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm9 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm9[4,5],ymm3[6,7]
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5],ymm3[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps (%rsp), %xmm5 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm9 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT:    # xmm9 = xmm5[3,3],mem[3,3]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT:    # xmm9 = xmm9[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm9[1,2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm9 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm8 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm8 = xmm5[3,3],mem[3,3]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm8 = xmm8[0,1,2],mem[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm8[1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm8 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[3,3,3,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm9 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2],ymm8[3,4],ymm9[5,6],ymm8[7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,1,2,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0],ymm0[1,2,3,4],ymm8[5,6,7]
-; AVX2-SLOW-NEXT:    vbroadcastss 176(%rdx), %ymm8
+; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm8 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1,2],ymm7[3,4],ymm8[5,6],ymm7[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[2,1,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0],ymm0[1,2,3,4],ymm7[5,6,7]
+; AVX2-SLOW-NEXT:    vbroadcastss 176(%rdx), %ymm7
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm0[3,1,2,0,7,5,6,4]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6],ymm9[7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6],ymm8[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm9 = ymm1[0],ymm12[0],ymm1[1],ymm12[1],ymm1[4],ymm12[4],ymm1[5],ymm12[5]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm1[0],ymm10[0],ymm1[1],ymm10[1],ymm1[4],ymm10[4],ymm1[5],ymm10[5]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5],ymm7[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm9 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT:    # xmm9 = xmm5[3,3],mem[3,3]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT:    # xmm9 = xmm9[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4,5,6,7]
-; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm9 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
-; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm12[2],ymm1[2],ymm12[3],ymm1[3],ymm12[6],ymm1[6],ymm12[7],ymm1[7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[3,3,3,3]
+; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm8 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm8 = xmm5[3,3],mem[3,3]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm8 = xmm8[0,1,2],mem[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1,2,3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm8 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm10[2],ymm1[2],ymm10[3],ymm1[3],ymm10[6],ymm1[6],ymm10[7],ymm1[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm9 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0],ymm9[1,2],ymm11[3,4],ymm9[5,6],ymm11[7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,1,2,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0],ymm0[1,2,3,4],ymm9[5,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm9 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[4],mem[4],ymm1[5],mem[5]
-; AVX2-SLOW-NEXT:    vpermilps $39, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    # ymm11 = mem[3,1,2,0,7,5,6,4]
-; AVX2-SLOW-NEXT:    vbroadcastss 208(%rdx), %ymm12
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6],ymm11[7]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3],ymm9[4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm8 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0],ymm8[1,2],ymm10[3,4],ymm8[5,6],ymm10[7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,1,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0],ymm0[1,2,3,4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm8 = ymm11[0],mem[0],ymm11[1],mem[1],ymm11[4],mem[4],ymm11[5],mem[5]
+; AVX2-SLOW-NEXT:    vpermilps $39, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT:    # ymm10 = mem[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vbroadcastss 208(%rdx), %ymm11
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6],ymm10[7]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3],ymm8[4,5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm11 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT:    # xmm11 = xmm1[3,3],mem[3,3]
-; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT:    # xmm11 = xmm11[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1,2,3],ymm9[4,5,6,7]
+; AVX2-SLOW-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm10 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm10 = xmm1[3,3],mem[3,3]
+; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT:    # xmm10 = xmm10[0,1,2],mem[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm10[1,2,3],ymm8[4,5,6,7]
 ; AVX2-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX2-SLOW-NEXT:    vmovaps %ymm9, 1440(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm8, 1440(%rax)
 ; AVX2-SLOW-NEXT:    vmovaps %ymm0, 1312(%rax)
-; AVX2-SLOW-NEXT:    vmovaps %ymm8, 1216(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm7, 1216(%rax)
 ; AVX2-SLOW-NEXT:    vmovaps %ymm2, 1088(%rax)
 ; AVX2-SLOW-NEXT:    vmovaps %ymm3, 992(%rax)
 ; AVX2-SLOW-NEXT:    vmovaps %ymm4, 864(%rax)
 ; AVX2-SLOW-NEXT:    vmovaps %ymm6, 768(%rax)
-; AVX2-SLOW-NEXT:    vmovaps %ymm13, 640(%rax)
-; AVX2-SLOW-NEXT:    vmovaps %ymm14, 544(%rax)
-; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vmovaps %ymm0, 416(%rax)
-; AVX2-SLOW-NEXT:    vmovaps %ymm10, 320(%rax)
-; AVX2-SLOW-NEXT:    vmovaps %ymm7, 192(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm12, 640(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm13, 544(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm15, 416(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm14, 320(%rax)
+; AVX2-SLOW-NEXT:    vmovaps %ymm9, 192(%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
@@ -9363,7 +9362,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-SLOW-NEXT:    vmovaps %ymm0, 1600(%rax)
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps %ymm0, 1568(%rax)
-; AVX2-SLOW-NEXT:    addq $2984, %rsp # imm = 0xBA8
+; AVX2-SLOW-NEXT:    addq $2968, %rsp # imm = 0xB98
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -9715,48 +9714,48 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vmovaps 224(%rdi), %xmm7
+; AVX2-FAST-NEXT:    vmovaps 224(%rdi), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps 224(%rsi), %xmm1
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[1,1,2,2]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm7[2],xmm2[3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2],xmm2[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,2,1]
 ; AVX2-FAST-NEXT:    vmovaps 224(%rcx), %xmm3
 ; AVX2-FAST-NEXT:    vmovaps 224(%rdx), %xmm6
-; AVX2-FAST-NEXT:    vinsertps {{.*#+}} xmm4 = zero,xmm6[1],xmm3[1],zero
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT:    vinsertps {{.*#+}} xmm5 = zero,xmm6[1],xmm3[1],zero
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1,2],ymm2[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 228(%r8), %ymm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT:    vmovaps 224(%r9), %xmm4
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm5 = xmm4[1,1,1,1]
-; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm5[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT:    vmovaps 224(%rax), %xmm2
-; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm6, %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0],ymm15[1],ymm5[2,3,4],ymm15[5],ymm5[6,7]
-; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vbroadcastss %xmm3, %xmm5
-; AVX2-FAST-NEXT:    vbroadcastss %xmm6, %xmm15
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm15[0],xmm5[0],xmm15[1],xmm5[1]
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
+; AVX2-FAST-NEXT:    vmovaps 224(%r9), %xmm7
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm7[1,1,1,1]
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm15[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovaps 224(%rax), %xmm15
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm15, %ymm5, %ymm5
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vbroadcastss %xmm3, %xmm2
+; AVX2-FAST-NEXT:    vbroadcastss %xmm6, %xmm5
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX2-FAST-NEXT:    vbroadcastf128 {{.*#+}} ymm10 = [0,1,2,2,0,1,2,2]
 ; AVX2-FAST-NEXT:    # ymm10 = mem[0,1,0,1]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm10, %ymm0
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm15[2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT:    vbroadcastsd 224(%r8), %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT:    vbroadcastss %xmm4, %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm15[5],ymm0[6,7]
-; AVX2-FAST-NEXT:    vbroadcastss %xmm2, %ymm2
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6],ymm0[7]
-; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm7[3,3],xmm1[3,3]
+; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm10, %ymm2
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT:    vbroadcastsd 224(%r8), %ymm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss %xmm7, %ymm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5],ymm2[6,7]
+; AVX2-FAST-NEXT:    vbroadcastss %xmm15, %ymm4
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6],ymm2[7]
+; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm10, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vmovaps 224(%r8), %ymm6
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,2,2,2]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 232(%rax), %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
@@ -9765,39 +9764,39 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-NEXT:    vmovaps 224(%rsi), %ymm1
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4],ymm2[5],ymm0[6,7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm0[2,2,2,2]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm0[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps 224(%rdx), %ymm3
 ; AVX2-FAST-NEXT:    vmovaps 224(%rcx), %ymm0
 ; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm15 = ymm3[1,1],ymm0[1,1],ymm3[5,5],ymm0[5,5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6],ymm7[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm15[5,6],ymm4[7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 240(%r8), %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm15[0],ymm7[1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm15[0],ymm4[1,2,3,4,5,6],ymm15[7]
 ; AVX2-FAST-NEXT:    vbroadcastss 240(%r9), %xmm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm15[1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm15[1],ymm4[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 240(%rax), %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2],ymm7[3,4,5,6,7]
-; AVX2-FAST-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm9[2],ymm8[2],ymm9[3],ymm8[3],ymm9[6],ymm8[6],ymm9[7],ymm8[7]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[3,3,3,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm15[2],ymm4[3,4,5,6,7]
+; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm4 = ymm9[2],ymm8[2],ymm9[3],ymm8[3],ymm9[6],ymm8[6],ymm9[7],ymm8[7]
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[3,3,3,3]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm11[2],ymm13[2],ymm11[3],ymm13[3],ymm11[6],ymm13[6],ymm11[7],ymm13[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm15[0,1,2],ymm7[3,4,5,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm15[0,1,2],ymm4[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 220(%r8), %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5],ymm7[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm15[5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 220(%r9), %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm15[6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 216(%rax), %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm15[0],ymm7[1,2,3,4,5,6],ymm15[7]
-; AVX2-FAST-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vbroadcastss 240(%rdx), %ymm7
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm15[0],ymm4[1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT:    vbroadcastss 240(%rdx), %ymm4
 ; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm0[3,1,2,0,7,5,6,4]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm15[0,1,2,3,4,5],ymm7[6],ymm15[7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm15[0,1,2,3,4,5],ymm4[6],ymm15[7]
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} ymm15 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm15[4,5],ymm7[6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm15[4,5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 236(%r8), %ymm15
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm15[1],ymm7[2,3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[2,2,3,3]
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm15[1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[2,2,3,3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3],ymm4[4,5,6,7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,2]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm3[2],ymm0[2],ymm3[3],ymm0[3],ymm3[6],ymm0[6],ymm3[7],ymm0[7]
@@ -10115,9 +10114,9 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vbroadcastss %xmm4, %xmm1
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm10, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
@@ -10128,7 +10127,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm1[4,5,6],ymm8[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm7[3,3],xmm0[3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm7[3,3]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm4[2],xmm5[2],xmm4[3],xmm5[3]
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm10, %ymm1
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
@@ -10442,31 +10441,31 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX2-FAST-PERLANE-LABEL: store_i32_stride7_vf64:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    subq $2984, %rsp # imm = 0xBA8
+; AVX2-FAST-PERLANE-NEXT:    subq $2968, %rsp # imm = 0xB98
 ; AVX2-FAST-PERLANE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rax), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r8), %xmm2
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r8), %xmm4
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r8), %xmm5
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r8), %xmm3
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r9), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r9), %xmm7
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r9), %xmm4
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %xmm9
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %xmm3
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %xmm10
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %xmm8
 ; AVX2-FAST-PERLANE-NEXT:    vinsertps {{.*#+}} xmm1 = zero,xmm8[1],xmm9[1],zero
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %xmm10
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %xmm7
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %xmm11
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -10479,15 +10478,15 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rax), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[1,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm11[1,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm10[2],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %xmm7
-; AVX2-FAST-PERLANE-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm7[1],xmm3[1],zero
+; AVX2-FAST-PERLANE-NEXT:    vinsertps {{.*#+}} xmm2 = zero,xmm7[1],xmm10[1],zero
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10666,11 +10665,11 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[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 64(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %ymm0
+; 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:    vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdx), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10806,13 +10805,13 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertps {{.*#+}} xmm14 = zero,xmm1[1],xmm0[1],zero
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1,2],ymm13[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 228(%r8), %ymm14
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3],ymm13[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm14 = xmm3[1,1,1,1]
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm14[4,5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm15, %ymm1, %ymm14
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 228(%r8), %ymm12
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm3[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3],xmm4[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
@@ -10828,37 +10827,37 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 232(%rax), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdi), %ymm10
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rsi), %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm4[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm10[1],ymm0[2,3,4],ymm10[5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdx), %ymm12
-; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rcx), %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm14 = ymm12[1,1],ymm2[1,1],ymm12[5,5],ymm2[5,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6],ymm13[7]
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 240(%r8), %ymm14
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0],ymm13[1,2,3,4,5,6],ymm14[7]
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 240(%r9), %xmm14
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 240(%rax), %ymm14
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm14[2],ymm13[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rsi), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm0[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdx), %ymm10
+; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rcx), %ymm1
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm10[1,1],ymm1[1,1],ymm10[5,5],ymm1[5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm13[5,6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 240(%r8), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0],ymm12[1,2,3,4,5,6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 240(%r9), %xmm13
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1],ymm12[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 240(%rax), %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1],ymm13[2],ymm12[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm9, %xmm13
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm8, %xmm14
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[0,1,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1],ymm13[2,3],ymm14[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm9, %xmm12
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm8, %xmm13
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm14[0],ymm15[0],ymm14[2],ymm15[2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm14[4,5,6],ymm13[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm13[0],ymm14[0],ymm13[2],ymm14[2]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[3,3],xmm5[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
@@ -10866,36 +10865,36 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 8(%rax), %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm1, %xmm5
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm12, %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm7, %xmm6
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm13[0],xmm9[1],xmm13[1]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm15[0],xmm9[1],xmm15[1]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm8[0],ymm6[2],ymm8[2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm14[3,3],xmm15[3,3]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm7[2],xmm1[2],xmm7[3],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm13[3,3],xmm14[3,3]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm7[2],xmm12[2],xmm7[3],xmm12[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm13[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm15[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 40(%rax), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -10905,28 +10904,28 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm13, %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm11, %xmm6
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm12[3,3],xmm9[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm11[2],xmm13[2],xmm11[3],xmm13[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 72(%rax), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
@@ -10936,9 +10935,9 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm14, %xmm6
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
@@ -10950,7 +10949,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm1[3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm12[3,3],xmm9[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
@@ -10962,41 +10961,41 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm12, %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm13, %xmm5
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm14, %xmm6
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm13, %xmm6
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm11[0],xmm9[0],xmm11[1],xmm9[1]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsp), %xmm0 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm11[3,3],xmm9[3,3]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm8[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm11[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 136(%rax), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm12, %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm13, %xmm5
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm14, %xmm6
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm13, %xmm6
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
@@ -11010,10 +11009,10 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
@@ -11029,39 +11028,39 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm13, %xmm6
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
+; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm8[3,3],xmm1[3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm11[2],xmm13[3],xmm11[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm9[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm12[0,1,2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 200(%rax), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[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:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm13[2],mem[2],ymm13[3],mem[3],ymm13[6],mem[6],ymm13[7],mem[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm6 = ymm9[2],ymm14[2],ymm9[3],ymm14[3],ymm9[6],ymm14[6],ymm9[7],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm6 = ymm13[2],ymm11[2],ymm13[3],ymm11[3],ymm13[6],ymm11[6],ymm13[7],ymm11[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 220(%r8), %ymm6
@@ -11072,17 +11071,17 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0],ymm5[1,2,3,4,5,6],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 240(%rdx), %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm2[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm1[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm6 = ymm10[0],ymm4[0],ymm10[1],ymm4[1],ymm10[4],ymm4[4],ymm10[5],ymm4[5]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm6 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 236(%r8), %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1],ymm5[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,2,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm10[2],ymm4[2],ymm10[3],ymm4[3],ymm10[6],ymm4[6],ymm10[7],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[6],ymm2[6],ymm4[7],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm6 = ymm12[2],ymm2[2],ymm12[3],ymm2[3],ymm12[6],ymm2[6],ymm12[7],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm6 = ymm10[2],ymm1[2],ymm10[3],ymm1[3],ymm10[6],ymm1[6],ymm10[7],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
@@ -11095,8 +11094,8 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm7[2,3],ymm6[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0],ymm3[1],ymm5[2,3,4],ymm3[5],ymm5[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm2[2],ymm12[2],ymm2[3],ymm12[3],ymm2[6],ymm12[6],ymm2[7],ymm12[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm10[2],ymm4[3],ymm10[3],ymm4[6],ymm10[6],ymm4[7],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm10[2],ymm1[3],ymm10[3],ymm1[6],ymm10[6],ymm1[7],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm4[2],ymm2[3],ymm4[3],ymm2[6],ymm4[6],ymm2[7],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -11112,28 +11111,28 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[1,1],ymm3[1,1],ymm7[5,5],ymm3[5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,1],ymm3[1,1],ymm8[5,5],ymm3[5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[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), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm14[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[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3,4],ymm1[5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 16(%rax), %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[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), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm15[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm10[1],ymm0[2,3,4],ymm10[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm6[1,1],ymm1[5,5],ymm6[5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm7[1,1],ymm1[5,5],ymm7[5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[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]
@@ -11145,14 +11144,14 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[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:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm15[1,1],ymm1[5,5],ymm15[5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm1[1,1],mem[1,1],ymm1[5,5],mem[5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[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]
@@ -11222,9 +11221,9 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[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:    vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm13[1,1],ymm0[5,5],ymm13[5,5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm12[1,1],ymm0[5,5],ymm12[5,5]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
@@ -11249,31 +11248,30 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    # xmm1 = xmm1[0,1,2],mem[3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm3[2],ymm7[2],ymm3[3],ymm7[3],ymm3[6],ymm7[6],ymm3[7],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm3[2],ymm8[2],ymm3[3],ymm8[3],ymm3[6],ymm8[6],ymm3[7],ymm8[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[3,3],ymm11[3,3],ymm8[7,7],ymm11[7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[3,3],ymm14[3,3],ymm9[7,7],ymm14[7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 48(%rdx), %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm10, %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm10[0],ymm12[0],ymm10[1],ymm12[1],ymm10[4],ymm12[4],ymm10[5],ymm12[5]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm10[0],ymm15[0],ymm10[1],ymm15[1],ymm10[4],ymm15[4],ymm10[5],ymm15[5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm6[2],mem[2],ymm6[3],mem[3],ymm6[6],mem[6],ymm6[7],mem[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm12[2],ymm2[2],ymm12[3],ymm2[3],ymm12[6],ymm2[6],ymm12[7],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm7[2],mem[2],ymm7[3],mem[3],ymm7[6],mem[6],ymm7[7],mem[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm15[2],ymm10[2],ymm15[3],ymm10[3],ymm15[6],ymm10[6],ymm15[7],ymm10[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -11283,24 +11281,23 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[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:    vblendps {{.*#+}} ymm15 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 80(%rdx), %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm15[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm9[0],ymm2[1],ymm9[1],ymm2[4],ymm9[4],ymm2[5],ymm9[5]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm6[0],ymm2[1],ymm6[1],ymm2[4],ymm6[4],ymm2[5],ymm6[5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm15[2],mem[2],ymm15[3],mem[3],ymm15[6],mem[6],ymm15[7],mem[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm2[2],ymm9[3],ymm2[3],ymm9[6],ymm2[6],ymm9[7],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm2[2],ymm6[3],ymm2[3],ymm6[6],ymm2[6],ymm6[7],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -11310,23 +11307,23 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 112(%rdx), %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[3,1,2,0,7,5,6,4]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 112(%rdx), %ymm2
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm1[4,5],ymm6[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm9[2],mem[2],ymm9[3],mem[3],ymm9[6],mem[6],ymm9[7],mem[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm1[1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -11334,92 +11331,90 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm9[0],ymm1[1,2],ymm9[3,4],ymm1[5,6],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3,4],ymm1[5,6],ymm8[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 144(%rdx), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm0[3,1,2,0,7,5,6,4]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5],ymm1[6],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3,4,5],ymm1[6],ymm8[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm9 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm9[4,5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5],ymm3[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsp), %xmm5 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm9 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # xmm9 = xmm5[3,3],mem[3,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # xmm9 = xmm9[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm9[1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm9 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm8 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm8 = xmm5[3,3],mem[3,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm8 = xmm8[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm8[1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm8 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm9 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2],ymm8[3,4],ymm9[5,6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,1,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0],ymm0[1,2,3,4],ymm8[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 176(%rdx), %ymm8
+; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm8 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1,2],ymm7[3,4],ymm8[5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[2,1,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0],ymm0[1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 176(%rdx), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm0[3,1,2,0,7,5,6,4]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6],ymm8[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm9 = ymm1[0],ymm12[0],ymm1[1],ymm12[1],ymm1[4],ymm12[4],ymm1[5],ymm12[5]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm1[0],ymm10[0],ymm1[1],ymm10[1],ymm1[4],ymm10[4],ymm1[5],ymm10[5]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5],ymm7[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm9 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # xmm9 = xmm5[3,3],mem[3,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # xmm9 = xmm9[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm9 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
-; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm12[2],ymm1[2],ymm12[3],ymm1[3],ymm12[6],ymm1[6],ymm12[7],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm8 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm8 = xmm5[3,3],mem[3,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm8 = xmm8[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm8 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm10[2],ymm1[2],ymm10[3],ymm1[3],ymm10[6],ymm1[6],ymm10[7],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm9 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0],ymm9[1,2],ymm11[3,4],ymm9[5,6],ymm11[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,1,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0],ymm0[1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm9 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[4],mem[4],ymm1[5],mem[5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps $39, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # ymm11 = mem[3,1,2,0,7,5,6,4]
-; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 208(%rdx), %ymm12
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6],ymm11[7]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3],ymm9[4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm8 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0],ymm8[1,2],ymm10[3,4],ymm8[5,6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,1,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0],ymm0[1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm8 = ymm11[0],mem[0],ymm11[1],mem[1],ymm11[4],mem[4],ymm11[5],mem[5]
+; AVX2-FAST-PERLANE-NEXT:    vpermilps $39, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # ymm10 = mem[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 208(%rdx), %ymm11
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3],ymm8[4,5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm11 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # xmm11 = xmm1[3,3],mem[3,3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    # xmm11 = xmm11[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1,2,3],ymm9[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm10 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm10 = xmm1[3,3],mem[3,3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT:    # xmm10 = xmm10[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm10[1,2,3],ymm8[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 1440(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, 1440(%rax)
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 1312(%rax)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, 1216(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm7, 1216(%rax)
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, 1088(%rax)
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm3, 992(%rax)
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm4, 864(%rax)
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm6, 768(%rax)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm13, 640(%rax)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm14, 544(%rax)
-; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 416(%rax)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm10, 320(%rax)
-; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm7, 192(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm12, 640(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm13, 544(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm15, 416(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm14, 320(%rax)
+; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, 192(%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
@@ -11508,7 +11503,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 1600(%rax)
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 1568(%rax)
-; AVX2-FAST-PERLANE-NEXT:    addq $2984, %rsp # imm = 0xBA8
+; AVX2-FAST-PERLANE-NEXT:    addq $2968, %rsp # imm = 0xB98
 ; AVX2-FAST-PERLANE-NEXT:    vzeroupper
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;

diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-3.ll
index 79ae689..67cfed6 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-3.ll
@@ -108,24 +108,24 @@ define void @store_i64_stride3_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride3_vf4:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm2
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdx), %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdi), %ymm4, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3],ymm3[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm4
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm5 = ymm1[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm2[2,3],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2],ymm4[3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd %ymm4, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdx), %ymm3, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm4 = ymm0[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2],ymm3[3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd %ymm3, 64(%rcx)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm0, 32(%rcx)
-; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%rcx)
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
@@ -223,42 +223,42 @@ define void @store_i64_stride3_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride3_vf8:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm2
-; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm4
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdx), %ymm5
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdx), %ymm1, %ymm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdi), %ymm7, %ymm7
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3],ymm6[4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT:    vmovapd 48(%rdi), %xmm7
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm8 = ymm3[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm5[2,3],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm2
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdx), %ymm3
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm4[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm5, %ymm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdx), %ymm5, %ymm5
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm5[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm6, %ymm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdx), %ymm6, %ymm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm7 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm3[2,3],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm7[0],ymm6[1],ymm7[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm8 = ymm0[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm2[2,3],ymm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm8[0],ymm7[1],ymm8[2],ymm7[3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdx), %ymm0, %ymm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdi), %ymm9, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm9
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm10 = ymm2[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm4[2,3],ymm10[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm10[0],ymm9[1],ymm10[2],ymm9[3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm2[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm3[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2],ymm2[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd %ymm9, 64(%rcx)
-; AVX1-ONLY-NEXT:    vmovaps %ymm8, 96(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm7, 160(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm0, 128(%rcx)
-; AVX1-ONLY-NEXT:    vmovaps %ymm6, (%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm1, 32(%rcx)
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd %ymm7, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm6, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm1, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm0, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, (%rcx)
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
@@ -444,91 +444,80 @@ define void @store_i64_stride3_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride3_vf16:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $40, %rsp
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm8
-; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm4
-; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm9
-; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdx), %ymm6
-; AVX1-ONLY-NEXT:    vmovapd 64(%rdx), %ymm10
-; AVX1-ONLY-NEXT:    vmovapd 96(%rdx), %ymm5
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdx), %ymm8, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdi), %ymm7, %ymm7
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm7[2,3],ymm1[4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 80(%rdi), %xmm7
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm10[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm11 = ymm9[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm10[2,3],ymm11[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm11[0],ymm7[1],ymm11[2],ymm7[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%rdx), %ymm2, %ymm11
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%rdi), %ymm12, %ymm12
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm12[2,3],ymm11[4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 48(%rdi), %xmm12
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm12 = ymm12[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm13 = ymm4[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm6[2,3],ymm13[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm13[0],ymm12[1],ymm13[2],ymm12[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdx), %ymm0, %ymm13
-; AVX1-ONLY-NEXT:    vmovapd %ymm0, %ymm12
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdi), %ymm14, %ymm14
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3],ymm13[4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm6
+; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm9
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdx), %ymm4
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdx), %ymm8
+; AVX1-ONLY-NEXT:    vmovapd 96(%rdx), %ymm3
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm5, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdx), %ymm5, %ymm5
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm0[2,3],ymm5[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm5[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm10, %ymm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%rdx), %ymm10, %ymm10
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0,1],ymm5[2,3],ymm10[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm10[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm11, %ymm10
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdx), %ymm11, %ymm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm11[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm12, %ymm11
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rdx), %ymm12, %ymm12
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3],ymm12[4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm12 = ymm12[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm13 = ymm6[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm8[2,3],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm12 = ymm13[0],ymm12[1],ymm13[2],ymm12[3]
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm13[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm14 = ymm2[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm4[2,3],ymm14[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm14[0],ymm13[1],ymm14[2],ymm13[3]
 ; AVX1-ONLY-NEXT:    vmovapd 112(%rdi), %xmm14
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm15 = ymm3[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm5[2,3],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm15 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm3[2,3],ymm15[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm15[0],ymm14[1],ymm15[2],ymm14[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rdi), %ymm15, %ymm15
-; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rdx), %ymm2, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm15[2,3],ymm1[4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm15
-; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm7
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3]
-; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm11 = ymm0[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm15[2,3],ymm11[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm11[0],ymm7[1],ymm11[2],ymm7[3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm8[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm15[1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm8 = ymm9[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm8 = ymm8[0,1],mem[2],ymm8[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm8[0],ymm10[1],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm4 = ymm4[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm12[2],ymm4[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0],ymm6[1],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm3[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vmovapd %ymm7, 64(%rcx)
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, 288(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm15[0,1],ymm9[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm7[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm9[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm15[1],ymm0[2],ymm15[3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm7 = ymm7[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2],ymm7[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0],ymm9[1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm6 = ymm6[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],mem[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0],ymm8[1],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm2[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],mem[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd %ymm0, 64(%rcx)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm14, 352(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm2, 320(%rcx)
-; AVX1-ONLY-NEXT:    vmovaps %ymm13, 96(%rcx)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, 160(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm4, 128(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm8, 224(%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 (%rsp), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm0, 32(%rcx)
-; AVX1-ONLY-NEXT:    addq $40, %rsp
+; AVX1-ONLY-NEXT:    vmovapd %ymm1, 320(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm13, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm2, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm6, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm12, 256(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm7, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm11, 288(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm10, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm5, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%rcx)
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
@@ -892,206 +881,186 @@ define void @store_i64_stride3_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride3_vf32:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $680, %rsp # imm = 0x2A8
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm3
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm0
+; AVX1-ONLY-NEXT:    subq $424, %rsp # imm = 0x1A8
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm3
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm12
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdx), %ymm13
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdx), %ymm4
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm6
-; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm5
-; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm4
-; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm9
-; AVX1-ONLY-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdx), %ymm8
-; AVX1-ONLY-NEXT:    vmovapd 64(%rdx), %ymm7
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdx), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdi), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[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:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm9[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm6[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm9[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm3[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm12[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdx), %ymm3, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdi), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm5[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm2[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm13[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%rdx), %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%rdi), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 80(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm4[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm7[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm10
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm10[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rdi), %ymm0, %ymm0
-; 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, 96(%rdx), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 96(%rdx), %ymm7
+; AVX1-ONLY-NEXT:    vmovapd 96(%rdx), %ymm4
 ; AVX1-ONLY-NEXT:    vmovapd 112(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm14
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm14[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm7[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm9
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm9[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%rdi), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 128(%rdi), %ymm12
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%rdx), %ymm12, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 128(%rdx), %ymm5
+; AVX1-ONLY-NEXT:    vmovapd 128(%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vmovapd 144(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 128(%rsi), %ymm10
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm10[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 160(%rdi), %ymm9
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%rdx), %ymm9, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 160(%rdx), %ymm4
-; AVX1-ONLY-NEXT:    vmovapd 176(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 160(%rsi), %ymm8
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm8[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%rdi), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 192(%rdi), %ymm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%rdx), %ymm6, %ymm11
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm0[2,3],ymm11[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 192(%rdx), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 208(%rdi), %xmm11
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 192(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm15 = ymm3[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm1[2,3],ymm15[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm15[0],ymm11[1],ymm15[2],ymm11[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 128(%rsi), %ymm8
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm6 = ymm8[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm2[2,3],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm6[0],ymm0[1],ymm6[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 160(%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 176(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 160(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm11 = ymm7[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm1[2,3],ymm11[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm11[0],ymm6[1],ymm11[2],ymm6[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rdi), %ymm11, %ymm11
-; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm15
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rdx), %ymm15, %ymm13
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm11[2,3],ymm13[4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 224(%rdx), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 240(%rdi), %xmm13
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm13[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 224(%rsi), %ymm2
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm11 = ymm2[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm0[2,3],ymm11[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0],ymm13[1],ymm11[2],ymm13[3]
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = ymm13[0,1],mem[2],ymm13[3]
-; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = ymm13[0],mem[1],ymm13[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = ymm13[0,1],mem[2],ymm13[3]
-; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = ymm13[0],mem[1],ymm13[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = ymm13[0,1],mem[2],ymm13[3]
-; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = ymm13[0],mem[1],ymm13[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm14 = ymm14[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm14 = ymm14[0,1],mem[2],ymm14[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm14[0],ymm7[1],ymm14[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 192(%rdx), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 208(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 192(%rsi), %ymm5
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm15 = ymm5[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm0[2,3],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm15[0],ymm6[1],ymm15[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovapd 224(%rdx), %ymm15
+; AVX1-ONLY-NEXT:    vmovapd 240(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 224(%rsi), %ymm3
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm11 = ymm3[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm15[2,3],ymm11[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0],ymm14[1],ymm11[2],ymm14[3]
+; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = mem[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm14[0,1],mem[2],ymm14[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm12 = ymm14[0],ymm12[1],ymm14[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = mem[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm14[0,1],mem[2],ymm14[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm14[0],ymm13[1],ymm14[2,3]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm10 = ymm10[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm12[2],ymm10[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm10[0],ymm5[1],ymm10[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1],mem[2],ymm10[3]
+; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm10 = ymm10[0],mem[1],ymm10[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm9 = ymm9[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],mem[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm9[0],ymm4[1],ymm9[2,3]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm8 = ymm8[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm8[0],ymm4[1],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm8[0,1],mem[2],ymm8[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm8[0],ymm2[1],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm7 = ymm7[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2],ymm7[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm7[0],ymm1[1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm5 = ymm5[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],mem[2],ymm5[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm5[0],ymm0[1],ymm5[2,3]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm3[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm3[0],ymm1[1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm2[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm15[2],ymm2[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],mem[2],ymm3[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0],ymm15[1],ymm3[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd %ymm11, 736(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm0, 704(%rcx)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm3, 704(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm6, 640(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm0, 608(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 640(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm1, 608(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm1, 512(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 576(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm2, 416(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm4, 512(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 352(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm4, 320(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 480(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm10, 224(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm5, 416(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm13, 128(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 384(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm12, 32(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 352(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm7, 320(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 576(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm13, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 480(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 384(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rcx)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%rcx)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%rcx)
-; AVX1-ONLY-NEXT:    addq $680, %rsp # imm = 0x2A8
+; AVX1-ONLY-NEXT:    addq $424, %rsp # imm = 0x1A8
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
@@ -1831,65 +1800,133 @@ define void @store_i64_stride3_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride3_vf64:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $1992, %rsp # imm = 0x7C8
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm3
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm0
+; AVX1-ONLY-NEXT:    subq $1480, %rsp # imm = 0x5C8
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm6
-; AVX1-ONLY-NEXT:    vmovupd %ymm6, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm5
-; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm4
-; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm9
-; AVX1-ONLY-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdx), %ymm8
-; AVX1-ONLY-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 64(%rdx), %ymm7
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdx), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdi), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[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:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 256(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 320(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 352(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 384(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 416(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[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:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm2
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm9[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm6[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm9[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdx), %ymm3, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdi), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdx), %ymm2
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm5[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%rdx), %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%rdi), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdx), %ymm2
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 80(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm4[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm7[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rdi), %ymm0, %ymm0
-; 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, 96(%rdx), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 96(%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 112(%rdi), %xmm0
@@ -1900,13 +1937,6 @@ define void @store_i64_stride3_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%rdi), %ymm0, %ymm0
-; 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, 128(%rdx), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 128(%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 144(%rdi), %xmm0
@@ -1917,358 +1947,257 @@ define void @store_i64_stride3_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%rdx), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 160(%rdx), %ymm2
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 160(%rdx), %ymm14
 ; AVX1-ONLY-NEXT:    vmovapd 176(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm14[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 160(%rsi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm14[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%rdi), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%rdx), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 192(%rdx), %ymm2
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 192(%rdx), %ymm12
 ; AVX1-ONLY-NEXT:    vmovapd 208(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm12[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 192(%rsi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm12[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rdi), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rdx), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 224(%rdx), %ymm2
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 224(%rdx), %ymm11
 ; AVX1-ONLY-NEXT:    vmovapd 240(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm11[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 224(%rsi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm11[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 256(%rdi), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 256(%rdx), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 256(%rdx), %ymm2
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 256(%rdx), %ymm9
 ; AVX1-ONLY-NEXT:    vmovapd 272(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm9[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 256(%rsi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm9[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%rdi), %ymm0, %ymm0
-; 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, 288(%rdx), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 288(%rdx), %ymm12
+; AVX1-ONLY-NEXT:    vmovapd 288(%rdx), %ymm8
 ; AVX1-ONLY-NEXT:    vmovapd 304(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm8[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 288(%rsi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm12[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 320(%rdi), %ymm0, %ymm0
-; 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, 320(%rdx), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 320(%rdx), %ymm10
+; AVX1-ONLY-NEXT:    vmovapd 320(%rdx), %ymm6
 ; AVX1-ONLY-NEXT:    vmovapd 336(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm10[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 320(%rsi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm10[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm6[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
-; 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, 352(%rdx), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 352(%rdx), %ymm8
+; AVX1-ONLY-NEXT:    vmovapd 352(%rdx), %ymm5
 ; AVX1-ONLY-NEXT:    vmovapd 368(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 352(%rsi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 384(%rdi), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 384(%rdx), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 384(%rdx), %ymm5
+; AVX1-ONLY-NEXT:    vmovapd 384(%rdx), %ymm4
 ; AVX1-ONLY-NEXT:    vmovapd 400(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 384(%rsi), %ymm14
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm14[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 384(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 416(%rdi), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 416(%rdi), %ymm13
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 416(%rdx), %ymm13, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 416(%rdx), %ymm4
+; AVX1-ONLY-NEXT:    vmovapd 416(%rdx), %ymm3
 ; AVX1-ONLY-NEXT:    vmovapd 432(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 416(%rsi), %ymm11
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm11[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 416(%rsi), %ymm13
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm13[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm3[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%rdi), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 448(%rdi), %ymm9
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%rdx), %ymm9, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 448(%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vmovapd 464(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 448(%rsi), %ymm7
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm7[0,0,3,2]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm2[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 448(%rsi), %ymm10
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm7 = ymm10[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm2[2,3],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm7[0],ymm0[1],ymm7[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rdi), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 480(%rdi), %ymm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rdx), %ymm6, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 480(%rdx), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 496(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 480(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm15 = ymm3[0,0,3,2]
+; AVX1-ONLY-NEXT:    vmovapd 496(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm7[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 480(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm15 = ymm7[0,0,3,2]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm1[2,3],ymm15[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm15[0],ymm0[1],ymm15[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilpd $1, (%rsp), %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1],mem[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1],mem[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1],mem[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1],mem[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, (%rsp) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1],mem[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1],mem[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1],mem[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1],mem[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1],mem[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm15 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm15 = ymm15[0,1],mem[2],ymm15[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm15[0,1],mem[2],ymm15[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm15[0],ymm14[1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = mem[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm15[0,1],mem[2],ymm15[3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm12 = ymm15[0],ymm12[1],ymm15[2,3]
 ; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm15 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm15 = ymm15[0,1],mem[2],ymm15[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm15[0],ymm10[1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm15[0,1],mem[2],ymm15[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm15[0],ymm11[1],ymm15[2,3]
 ; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm15 = mem[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm15 = ymm15[0,1],mem[2],ymm15[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm15[0,1],mem[2],ymm15[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm15[0],ymm9[1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = mem[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm15[0,1],mem[2],ymm15[3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm15[0],ymm8[1],ymm15[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm14 = ymm14[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd $4, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm14 = ymm14[0,1],mem[2],ymm14[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm14[0],ymm5[1],ymm14[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm11 = ymm11[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm13[2],ymm11[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm11[0],ymm4[1],ymm11[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = mem[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm15[0,1],mem[2],ymm15[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm15[0],ymm6[1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = mem[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm15[0,1],mem[2],ymm15[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm15[0],ymm5[1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm15 = mem[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm15[0,1],mem[2],ymm15[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm15[0],ymm4[1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm13 = ymm13[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm13[0,1],mem[2],ymm13[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm13[0],ymm3[1],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm10 = ymm10[1,0,2,2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1],mem[2],ymm10[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm10[0],ymm2[1],ymm10[2,3]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm7 = ymm7[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm9[2],ymm7[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm7[0],ymm2[1],ymm7[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm3[1,0,2,2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm3[0],ymm1[1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm3, 1504(%rcx)
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2],ymm7[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm7[0],ymm1[1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 1504(%rcx)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm1, 1472(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, 1440(%rcx)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 1408(%rcx)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm2, 1376(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, 1344(%rcx)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 1312(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm4, 1280(%rcx)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, 1248(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm3, 1280(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 1216(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm5, 1184(%rcx)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, 1152(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm4, 1184(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 1120(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm8, 1088(%rcx)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, 1056(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm5, 1088(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 1024(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm10, 992(%rcx)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, 960(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm6, 992(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 928(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm12, 896(%rcx)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, 864(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm8, 896(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 832(%rcx)
-; AVX1-ONLY-NEXT:    vmovapd %ymm0, 800(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm9, 800(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 736(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm11, 704(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 640(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm12, 608(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 544(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm14, 512(%rcx)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 448(%rcx)
+; AVX1-ONLY-NEXT:    vmovapd %ymm0, 416(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 768(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 352(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 736(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 320(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 704(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 640(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 608(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 576(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 512(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1440(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 480(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1344(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1248(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 416(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1152(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 384(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1056(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 352(%rcx)
-; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 320(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 960(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 864(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 768(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 576(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 480(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 384(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%rcx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%rcx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%rcx)
-; AVX1-ONLY-NEXT:    addq $1992, %rsp # imm = 0x7C8
+; AVX1-ONLY-NEXT:    addq $1480, %rsp # imm = 0x5C8
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;

diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-6.ll
index 825abed..c94df69 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-6.ll
@@ -185,33 +185,33 @@ define void @store_i64_stride6_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm3[0,1],ymm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm8[0],ymm5[1],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm8
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm8
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r9), %ymm8, %ymm9
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r8), %ymm11
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1],ymm9[2,3],ymm10[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm4[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm9
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm10 = xmm9[1],xmm8[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r9), %ymm10, %ymm11
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r8), %ymm12
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm12[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm11[2,3],ymm10[4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],mem[0],ymm2[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm11[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm4[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm10[0],ymm0[2],ymm10[3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm11[0],ymm0[2],ymm11[3]
 ; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm1
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 24(%r8), %ymm3
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm7[0],xmm6[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm4
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm4[0],xmm8[0]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm9[0],xmm8[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm4, 16(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, (%rax)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm0, 128(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm9, 64(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm2, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm10, 64(%rax)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm5, 32(%rax)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm1, 160(%rax)
 ; AVX1-ONLY-NEXT:    vzeroupper
@@ -418,89 +418,93 @@ define void @store_i64_stride6_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride6_vf8:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm9
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm13
-; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm11
-; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm12
-; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm14
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm11
+; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm13
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm4
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm3
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm12[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm11[0,1],ymm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm5
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm5[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm14[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm6[0],ymm2[1],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm7
-; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm8
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r9), %ymm8, %ymm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r8), %ymm15
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm15[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm6[2,3],ymm10[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r8), %ymm15
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm15[4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r9), %ymm7, %ymm15
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm15[2,3],ymm10[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r9), %ymm2, %ymm8
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r8), %ymm9
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm8[2,3],ymm2[4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm9
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm9[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r9), %ymm8, %ymm10
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r8), %ymm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm14[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm10[2,3],ymm8[4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm10 = mem[0,0]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm15 = xmm14[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm13[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm15[0],ymm10[1],ymm15[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm15
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm12 = ymm12[1],ymm15[1],ymm12[3],ymm15[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm13[2,3],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %ymm0
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm0[2,3],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm12 = ymm12[0],ymm13[0],ymm12[2],ymm13[3]
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm13
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm15
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm13[1],ymm15[1],ymm13[3],ymm15[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm14[2,3],ymm13[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %ymm14
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm14[2,3],ymm15[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm13 = ymm13[0],ymm15[0],ymm13[2],ymm15[3]
-; AVX1-ONLY-NEXT:    vmovapd (%r9), %ymm15
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm9 = ymm9[1],ymm11[1],ymm9[3],ymm11[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm15[2,3],ymm11[2,3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm12[2,3],ymm9[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm11 = ymm9[0],ymm11[0],ymm9[2],ymm11[3]
-; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm9
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm9 = xmm9[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 56(%r8), %ymm12
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm12[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1,2],ymm14[3]
-; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm12
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm14
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm14
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm14 = xmm14[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 24(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm14[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm11[2,3],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vmovapd (%r9), %ymm1
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm1[2,3],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm13 = ymm11[0],ymm13[0],ymm11[2],ymm13[3]
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm11
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm11 = xmm11[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 56(%r8), %ymm15
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm15
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 24(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm15[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 16(%rcx), %ymm15
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm5
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm5[0],xmm8[0]
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm1[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm7[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm14[0],xmm4[0]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm7[0],xmm6[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm9[0],xmm5[0]
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, 16(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, 208(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 16(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, (%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, 208(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm4, 192(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm11, 128(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm13, 320(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm10, 64(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm6, 256(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm14, 96(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm0, 160(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm2, 224(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm12, 288(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm9, 352(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm13, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm12, 320(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm1, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm10, 224(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm11, 352(%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 {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
 ; AVX1-ONLY-NEXT:    vzeroupper
@@ -1032,200 +1036,206 @@ define void @store_i64_stride6_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride6_vf16:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $376, %rsp # imm = 0x178
-; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm1
+; AVX1-ONLY-NEXT:    subq $456, %rsp # imm = 0x1C8
+; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm13
 ; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 64(%r8), %ymm4
-; AVX1-ONLY-NEXT:    vmovapd 96(%r8), %ymm7
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm5
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm5
 ; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm13[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm6
 ; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm5[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm1[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm6[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm10
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm10[1],xmm8[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm11
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm13
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm13[1],xmm11[1]
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %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:    vinsertf128 $1, (%r9), %ymm1, %ymm2
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r8), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm2[1],xmm4[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm14
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm14[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r9), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm9[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovapd 64(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm1[0,1],ymm2[2,3]
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm15
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r9), %ymm15, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm7[1],xmm8[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 72(%r8), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r9), %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vmovapd (%r9), %ymm12
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm12[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r8), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm5
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r9), %ymm5, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm2
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %ymm6
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm6[2,3],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm5[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovapd 96(%r8), %ymm4
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm2
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm10 = xmm2[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 104(%r8), %ymm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm10[0,1,2,3],ymm11[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r9), %ymm10, %ymm10
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm11
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm10[1],ymm11[1],ymm10[3],ymm11[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm13[2,3],ymm10[2,3]
+; AVX1-ONLY-NEXT:    vmovapd (%r9), %ymm15
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm15[2,3],ymm11[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm10 = ymm10[0],ymm11[0],ymm10[2],ymm11[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm11
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm10[1],ymm11[1],ymm10[3],ymm11[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm10[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %ymm10
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm10[2,3],ymm11[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm11[0],ymm0[2],ymm11[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 72(%r8), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm3
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r9), %ymm3, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 64(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm4[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm11
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm11[1],ymm0[3],ymm11[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 64(%r9), %ymm0
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm0[2,3],ymm11[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm11[0],ymm1[2],ymm11[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 104(%r8), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm2
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r9), %ymm2, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3],ymm1[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm4
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm7[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm11
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm11[1],ymm1[3],ymm11[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 96(%r9), %ymm1
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm9 = ymm7[0],ymm4[0],ymm7[2],ymm4[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm1[2,3],ymm11[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm4 = ymm4[0],ymm11[0],ymm4[2],ymm11[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, (%rsp) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm4
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 16(%rcx), %ymm7
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 16(%rcx), %ymm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm11[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm4
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm4[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 24(%r8), %ymm7
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 24(%r8), %ymm11
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm11[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1,2],ymm15[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm4
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm7
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm4[0,1,2,3,4,5],ymm11[6,7]
 ; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm4
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm4[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 56(%r8), %ymm7
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm4[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 56(%r8), %ymm15
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm4[0,1,2],ymm10[3]
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm4
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 80(%rcx), %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm4[0,1,2,3,4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovapd 80(%rdx), %xmm4
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm4[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 88(%r8), %ymm12
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm12[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm12
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm12
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm12 = xmm12[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 120(%r8), %ymm14
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm12 = ymm12[0,1],ymm14[2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 80(%rcx), %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdx), %xmm15
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 88(%r8), %ymm13
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm15[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm13[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm15
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 120(%r8), %ymm12
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm12[0,1,2],ymm1[3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm10 = xmm10[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm12
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm12[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm12 = xmm12[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm14
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm14[0],xmm5[0]
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm11 = xmm13[0],xmm11[0]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm13
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm13[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm13 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm14
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm14[0],xmm15[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm9 = xmm9[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm7[0],xmm8[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = xmm8[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm14[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm5[0],xmm6[0]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm3 = xmm3[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm6 = xmm6[0],mem[0]
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps %xmm14, 16(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm13, (%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, 16(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, (%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, 592(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm11, 576(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, 208(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm12, 192(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, 400(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, 384(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm9, 704(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm2, 640(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 576(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, 208(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, 400(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm9, 384(%rax)
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm2, 704(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm2, 512(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm2, 448(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm2, 320(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm2, 256(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm2, 128(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm2, 64(%rax)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm1, 736(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm13, 672(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 640(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 608(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm0, 544(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm4, 480(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 608(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm4, 544(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm6, 480(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 416(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm7, 352(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm8, 288(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm10, 352(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm11, 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 {{[-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
@@ -1233,8 +1243,10 @@ define void @store_i64_stride6_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; 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:    addq $376, %rsp # imm = 0x178
+; AVX1-ONLY-NEXT:    addq $456, %rsp # imm = 0x1C8
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
@@ -2250,459 +2262,471 @@ define void @store_i64_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride6_vf32:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $1576, %rsp # imm = 0x628
-; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm4
-; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm2
-; AVX1-ONLY-NEXT:    vmovapd 64(%r8), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 96(%r8), %ymm5
+; AVX1-ONLY-NEXT:    subq $1592, %rsp # imm = 0x638
+; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm4
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm6
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm7
-; AVX1-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm6[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm9
+; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm1[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r9), %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r8), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm8[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm2[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm11[1],xmm9[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm7[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm1[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm10[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 64(%r8), %ymm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 72(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 96(%r8), %ymm5
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm5[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 104(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm14
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm15[1],xmm14[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 128(%r8), %ymm8
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 128(%r8), %ymm3
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rcx), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 136(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vmovapd 160(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm6 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0],ymm6[1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm6
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm6[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm6
-; AVX1-ONLY-NEXT:    vmovapd 192(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm3[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 160(%r8), %ymm13
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm13[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 168(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 192(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm7[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 200(%r8), %ymm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm0[2,3],ymm7[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm7
 ; AVX1-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm6
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm6[1],xmm7[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT:    vmovapd 224(%r8), %ymm12
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm12[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r8), %ymm7
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm7[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm7
+; AVX1-ONLY-NEXT:    vmovapd 224(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm0[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm8 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm7[1],xmm8[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 232(%r8), %ymm12
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm7[0,1,2,3],ymm12[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%r9), %ymm7, %ymm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm12[0,1],ymm7[2,3],ymm12[4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r9), %ymm7, %ymm7
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3],ymm6[4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm6
-; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm9
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm6 = ymm6[1],ymm9[1],ymm6[3],ymm9[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm12
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm7 = ymm7[1],ymm12[1],ymm7[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm7[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd (%r9), %ymm7
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm7[2,3],ymm9[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm4 = ymm4[0],ymm6[0],ymm4[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm7[2,3],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm12[0],ymm1[2],ymm12[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm12
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm12[1],ymm1[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm1[2,3],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm4 = ymm4[0],ymm12[0],ymm4[2],ymm12[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r8), %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm6
-; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r9), %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3],ymm4[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm4
-; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm6
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm6[1],ymm4[3],ymm6[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %ymm15
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm15[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[2],ymm4[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 72(%r8), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm4
-; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r9), %ymm4, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3],ymm2[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 64(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm4
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm4[1],ymm2[3],ymm4[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 64(%r9), %ymm13
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm13[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm12
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm12[1],ymm4[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm6[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 64(%r9), %ymm6
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm6[2,3],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm4 = ymm4[0],ymm12[0],ymm4[2],ymm12[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm12
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm12[1],ymm4[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 96(%r9), %ymm4
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm4[2,3],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm12[0],ymm5[2],ymm12[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 128(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovapd 128(%rsi), %ymm12
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm5[1],ymm12[1],ymm5[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm3[2,3],ymm5[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 128(%r9), %ymm3
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm3[2,3],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm12[0],ymm5[2],ymm12[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 160(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovapd 160(%rsi), %ymm12
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm5[1],ymm12[1],ymm5[3],ymm12[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm13[2,3],ymm5[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 160(%r9), %ymm5
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm5[2,3],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm8 = ymm13[0],ymm12[0],ymm13[2],ymm12[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 192(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vmovapd 192(%rsi), %ymm13
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm12 = ymm12[1],ymm13[1],ymm12[3],ymm13[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm2[2,3],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 192(%r9), %ymm2
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm2[2,3],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm8 = ymm12[0],ymm13[0],ymm12[2],ymm13[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vmovapd 224(%rsi), %ymm13
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm12 = ymm12[1],ymm13[1],ymm12[3],ymm13[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm0[2,3],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 224(%r9), %ymm0
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm0[2,3],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm8 = ymm12[0],ymm13[0],ymm12[2],ymm13[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 16(%rcx), %ymm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm12[0,1,2,3,4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm12
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm12 = xmm12[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 24(%r8), %ymm13
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm12[0,1,2],ymm7[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm12
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm7[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 56(%r8), %ymm12
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm7[0,1,2],ymm1[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 104(%r8), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r9), %ymm2, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 80(%rcx), %ymm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm2
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 96(%r9), %ymm10
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm10[2,3],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 88(%r8), %ymm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm6[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 136(%r8), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rcx), %xmm11
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%r9), %ymm11, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 128(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 128(%rsi), %ymm2
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 128(%r9), %ymm8
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm8[2,3],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 120(%r8), %ymm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 168(%r8), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %xmm14
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%r9), %ymm14, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 144(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 144(%rcx), %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 160(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 160(%rsi), %ymm4
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 160(%r9), %ymm2
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 200(%r8), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd 144(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 152(%r8), %ymm4
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 176(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 176(%rcx), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%r9), %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 192(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 192(%rsi), %ymm4
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm4[1],ymm0[3],ymm4[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 192(%r9), %ymm1
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 232(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %xmm9
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%r9), %ymm9, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3],ymm0[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 224(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm3[1],ymm0[3],ymm3[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm12[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 224(%r9), %ymm0
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm0[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm3 = ymm5[0],ymm3[0],ymm5[2],ymm3[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 16(%rcx), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 24(%r8), %ymm5
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm7[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 56(%r8), %ymm5
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm15[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 80(%rcx), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 80(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 88(%r8), %ymm5
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 120(%r8), %ymm5
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm10[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 144(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 144(%rcx), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 144(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 152(%r8), %ymm5
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 176(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 176(%rcx), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 176(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 184(%r8), %ymm5
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 208(%rcx), %ymm3
-; 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:    vmovapd 208(%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm2[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 216(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovapd 176(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 184(%r8), %ymm3
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm5[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 240(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 240(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 208(%rcx), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 240(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd 208(%rdx), %xmm1
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 248(%r8), %ymm2
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 216(%r8), %ymm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 240(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 240(%rcx), %ymm5
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 240(%rdx), %xmm5
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm5[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 248(%r8), %ymm13
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm5[0,1,2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm15[0],xmm14[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm14 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm0[0],xmm10[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm11 = xmm11[0],xmm9[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm9 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm8 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm7 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd (%rsp), %xmm0, %xmm6 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm6 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm11 = xmm0[0],xmm11[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm5 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm15
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm15 = xmm15[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm4 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm13
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm13 = xmm13[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm3 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm12
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm12 = xmm12[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm2 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm10 = xmm10[0],xmm14[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm1 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm14
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm14[0],xmm9[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm14 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm8
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm8 = xmm8[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm7
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm7 = xmm7[0],mem[0]
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps %xmm7, 16(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, 1168(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm14, 1152(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm9, 1360(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, 1344(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, 976(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, 960(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm12, 592(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, 576(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm13, 208(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm4, 192(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm15, 400(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, 384(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm11, 784(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, 768(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, 16(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, 1168(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, 1152(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, 1360(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 1344(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, 976(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, 960(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 592(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm9, 576(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm10, 208(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm11, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, 400(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm13, 384(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm14, 784(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, 768(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1472(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1408(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1280(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1216(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1088(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1024(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 896(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 832(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 704(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 640(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 512(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%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, 256(%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, 64(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1504(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1440(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1408(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1376(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1312(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1248(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1216(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1184(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1120(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1056(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1024(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 992(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 928(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 864(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 832(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 800(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 736(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 640(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 608(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 480(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 416(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 352(%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 {{[-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, 160(%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:    addq $1576, %rsp # imm = 0x628
+; AVX1-ONLY-NEXT:    addq $1592, %rsp # imm = 0x638
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
@@ -4793,50 +4817,89 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-LABEL: store_i64_stride6_vf64:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    subq $3464, %rsp # imm = 0xD88
-; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %ymm7
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm4
-; AVX1-ONLY-NEXT:    vmovapd 64(%r8), %ymm14
-; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm5
+; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm6
+; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm7
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm6
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps (%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:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r9), %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r8), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm6[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm14[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 64(%r8), %ymm10
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm10[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 72(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 96(%r8), %ymm8
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 104(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4844,554 +4907,554 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 128(%r8), %ymm7
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm7[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 128(%r8), %ymm14
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm14[0,1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 136(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 160(%r8), %ymm6
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm2[1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[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(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 168(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 192(%r8), %ymm2
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm3[1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 192(%r8), %ymm15
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 200(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 224(%r8), %ymm15
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm15[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm3[1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 256(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 224(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[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(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 256(%r8), %ymm13
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm13[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm3[1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 288(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 232(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 288(%r8), %ymm11
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm3[1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 320(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 256(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vmovapd 320(%r8), %ymm1
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm1[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm8 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm3[0],ymm8[1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 352(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm8
-; AVX1-ONLY-NEXT:    vmovapd 352(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm3[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm10 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm8[0],ymm10[1],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 264(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 256(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 288(%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-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:    vmovaps 384(%rdi), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT:    vmovapd 384(%r8), %ymm12
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm12[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm10 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm8[0],ymm10[1],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 416(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 288(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[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(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT:    vmovaps 416(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 296(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm8[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm10 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm10[2,3],ymm8[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 320(%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 320(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[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(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 320(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT:    vmovapd 448(%r8), %ymm10
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm10[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm8[0],ymm9[1],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 328(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 320(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 352(%r8), %ymm11
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm11[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 480(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 352(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT:    vmovaps 480(%r8), %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[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm9[2,3],ymm8[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r8), %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm0
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r9), %ymm0, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm9[2,3],ymm8[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 360(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 352(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm8
-; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm9
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm8[1],ymm9[1],ymm8[3],ymm9[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vmovapd (%r9), %ymm0
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm0[2,3],ymm9[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm5[0],ymm9[0],ymm5[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovaps 384(%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 384(%r8), %ymm9
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm9[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r8), %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm9[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm0
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r9), %ymm0, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm9[2,3],ymm5[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 384(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 392(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 384(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm5
-; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm9
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm5[1],ymm9[1],ymm5[3],ymm9[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %ymm5
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm5[2,3],ymm9[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm4[0],ymm9[0],ymm4[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovaps 416(%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-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:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 416(%r8), %ymm12
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm12[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 72(%r8), %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm9[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm0
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r9), %ymm0, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm9[2,3],ymm4[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 416(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 424(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 416(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 64(%rdi), %ymm4
-; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm9
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm9[1],ymm4[3],ymm9[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm14[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 64(%r9), %ymm14
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm14[2,3],ymm9[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm4[0],ymm9[0],ymm4[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovaps 448(%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-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:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 448(%r8), %ymm13
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm13[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 104(%r8), %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm9[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm0
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r9), %ymm0, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm9[2,3],ymm4[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 448(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 456(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%r9), %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm4
-; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm9
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm9[1],ymm4[3],ymm9[3]
-; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 96(%r9), %ymm4
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm4[2,3],ymm9[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovaps 480(%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 480(%r8), %ymm3
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm2[1],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 136(%r8), %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rcx), %xmm8
-; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%r9), %ymm8, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm9[2,3],ymm0[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 480(%rcx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 488(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%r9), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 128(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 128(%rsi), %ymm9
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm9[1],ymm0[3],ymm9[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 128(%r9), %ymm7
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm7[2,3],ymm9[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd (%r9), %ymm4
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm4[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 168(%r8), %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %xmm8
-; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%r9), %ymm8, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm9[2,3],ymm0[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 160(%rdi), %ymm9
-; AVX1-ONLY-NEXT:    vmovapd 160(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm9 = ymm9[1],ymm0[1],ymm9[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm6[2,3],ymm9[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 160(%r9), %ymm8
-; AVX1-ONLY-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm6[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %ymm5
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm5[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm10[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 64(%r9), %ymm6
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm6[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm7[1],ymm0[3],ymm7[3]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm8[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 96(%r9), %ymm10
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm10[2,3],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm7[0],ymm0[2],ymm7[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 128(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 128(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm7[1],ymm0[3],ymm7[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 128(%r9), %ymm14
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm14[2,3],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm7[0],ymm0[2],ymm7[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 160(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 160(%rsi), %ymm8
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm8[1],ymm0[3],ymm8[3]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 160(%r9), %ymm7
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm7[2,3],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm8[0],ymm0[2],ymm8[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 200(%r8), %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm6
-; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%r9), %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 192(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 192(%rsi), %ymm6
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm6[1],ymm0[3],ymm6[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 192(%r9), %ymm8
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm8[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovapd 192(%rsi), %ymm8
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm8[1],ymm0[3],ymm8[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 192(%r9), %ymm15
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm15[2,3],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm8[0],ymm0[2],ymm8[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 232(%r8), %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %xmm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%r9), %ymm2, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 224(%rsi), %ymm6
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm6[1],ymm0[3],ymm6[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 224(%r9), %ymm15
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm15[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovapd 224(%rsi), %ymm8
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm8[1],ymm0[3],ymm8[3]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 224(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm1[2,3],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm8[0],ymm0[2],ymm8[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 264(%r8), %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 256(%rcx), %xmm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 256(%r9), %ymm2, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 256(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 256(%rsi), %ymm6
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm6[1],ymm0[3],ymm6[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 256(%r9), %ymm13
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm13[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovapd 256(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovapd 256(%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm8[1],ymm0[1],ymm8[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 256(%r9), %ymm8
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm8[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 296(%r8), %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 288(%rcx), %xmm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%r9), %ymm2, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 288(%rdi), %ymm6
-; AVX1-ONLY-NEXT:    vmovapd 288(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm6 = ymm6[1],ymm0[1],ymm6[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm11[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 288(%r9), %ymm6
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm6[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm11[0],ymm0[0],ymm11[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 288(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 288(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 288(%r9), %ymm0
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm2[0],ymm1[0],ymm2[2],ymm1[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 328(%r8), %ymm11
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm11[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 320(%rcx), %xmm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 320(%r9), %ymm2, %ymm11
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm11[2,3],ymm0[4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 320(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 320(%rsi), %ymm11
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm11[1],ymm0[3],ymm11[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 320(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 320(%r9), %ymm2
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm2[2,3],ymm11[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm11[0],ymm1[2],ymm11[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 360(%r8), %ymm11
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm11[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 352(%rcx), %xmm0
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 352(%r9), %ymm0, %ymm11
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm11[2,3],ymm1[4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 352(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 352(%rsi), %ymm11
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm11[1],ymm1[3],ymm11[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 352(%r9), %ymm9
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm9[2,3],ymm11[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm3[0],ymm11[0],ymm3[2],ymm11[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 392(%r8), %ymm11
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm11[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 384(%rcx), %xmm0
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 384(%r9), %ymm0, %ymm11
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm11[2,3],ymm3[4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 384(%rdi), %ymm3
-; AVX1-ONLY-NEXT:    vmovapd 384(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm12[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 384(%r9), %ymm11
+; AVX1-ONLY-NEXT:    vmovapd 352(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 352(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 424(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 416(%rcx), %xmm3
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 416(%r9), %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3],ymm0[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 416(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 416(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm3[1],ymm0[3],ymm3[3]
-; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 416(%r9), %ymm12
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm12[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[3]
+; AVX1-ONLY-NEXT:    vmovapd 352(%r9), %ymm11
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm11[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 456(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 448(%rcx), %xmm0
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%r9), %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm3[2,3],ymm1[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 448(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 448(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm3[1],ymm0[3],ymm3[3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm10[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 448(%r9), %ymm10
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm10[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[3]
+; AVX1-ONLY-NEXT:    vmovapd 384(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 384(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm9[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 384(%r9), %ymm0
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 488(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 480(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%r9), %ymm1, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3],ymm0[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 480(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 480(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm3[1],ymm0[3],ymm3[3]
-; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 480(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm9[0],ymm1[0],ymm9[2],ymm1[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 416(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 416(%rsi), %ymm9
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm9[1],ymm1[3],ymm9[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm12[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 416(%r9), %ymm12
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm12[2,3],ymm9[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm9[0],ymm1[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 448(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 448(%rsi), %ymm9
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm9[1],ymm1[3],ymm9[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm13[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 448(%r9), %ymm13
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm13[2,3],ymm9[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm9[0],ymm1[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 480(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 480(%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm3[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 480(%r9), %ymm9
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm9[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 16(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 24(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 56(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 80(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 80(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 88(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm14[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 120(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 144(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 144(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 144(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 152(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm7[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 176(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 176(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 176(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 184(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 208(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 208(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 216(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 240(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 240(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm5[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 80(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 88(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 120(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm10[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 144(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 144(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 144(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 152(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 176(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 176(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 176(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 184(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm7[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 208(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 208(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 216(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm15[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 240(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 240(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 240(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 248(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 272(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 272(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 272(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 280(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 304(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 304(%rcx), %ymm2
+; 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:    vmovapd 240(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 248(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 272(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 272(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 304(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 312(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm1[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 272(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 280(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm13[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 304(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 304(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 336(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 336(%rcx), %ymm2
+; 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:    vmovapd 304(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 312(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm6[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 336(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 336(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 336(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 344(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm1[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 336(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 344(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 368(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 368(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 368(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; 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:    vmovapd 368(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 368(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 376(%r8), %ymm2
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm11[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 400(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 400(%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 400(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 400(%rcx), %ymm2
+; 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:    vmovaps 400(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 408(%r8), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm1[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 400(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 408(%r8), %ymm1
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm11[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 432(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
@@ -5414,7 +5477,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 472(%r8), %ymm1
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm10[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm13[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 496(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
@@ -5422,18 +5485,17 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vbroadcastsd 496(%rcx), %ymm1
 ; 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 496(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovapd 496(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 504(%r8), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovupd %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:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 256(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5441,7 +5503,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5449,15 +5511,15 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5465,15 +5527,15 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -5481,7 +5543,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5489,75 +5551,75 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm15 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm14 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm13 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm12 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm10 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm9 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm8 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm8 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 288(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm7 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm7 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm6 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 352(%rdx), %xmm15
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm15 = xmm15[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm5 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 320(%rdx), %xmm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm14 = xmm14[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm4 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 416(%rdx), %xmm13
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm13 = xmm13[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm3 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 480(%rdx), %xmm12
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm12 = xmm12[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm2 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 448(%rdx), %xmm11
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm11 = xmm11[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm1 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm10 = xmm10[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm9
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm9 = xmm9[0],mem[0]
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps %xmm9, 16(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, 2320(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, 2304(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm11, 2704(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, 2688(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm12, 2896(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, 2880(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm13, 2512(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm4, 2496(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm14, 1936(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, 1920(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm15, 2128(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, 2112(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm7, 1744(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, 1728(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, 16(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, 2320(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, 2304(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, 2704(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 2688(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, 2896(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, 2880(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 2512(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm9, 2496(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm10, 1936(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm11, 1920(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, 2128(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm13, 2112(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm14, 1744(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, 1728(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 1168(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -5593,162 +5655,162 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3008(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2944(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2816(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2752(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2624(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2560(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2432(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2368(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2240(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2176(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2048(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1984(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1856(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1792(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1664(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1600(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1472(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1408(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1280(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1216(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1088(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1024(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 896(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 832(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 704(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 640(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 512(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%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, 256(%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, 64(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3040(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2976(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2944(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2912(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2848(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2784(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2752(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2720(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2656(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2592(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2560(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2528(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2464(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2400(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2368(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2336(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2272(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2208(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2176(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2144(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2080(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2016(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1984(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1952(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1888(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1824(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1792(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1760(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1696(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1632(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1600(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1568(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1504(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1440(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1408(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1376(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1312(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1248(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1216(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1184(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1120(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1056(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1024(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 992(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 928(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 864(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 832(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 800(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 736(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 640(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 608(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 480(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 416(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 352(%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 {{[-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, 160(%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:    addq $3464, %rsp # imm = 0xD88
 ; AVX1-ONLY-NEXT:    vzeroupper

diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
index eadc265..923d658 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
@@ -206,26 +206,26 @@ define void @store_i64_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm6 = ymm3[1],ymm6[1],ymm3[3],ymm6[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm6, %ymm7
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm8[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm9, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm7
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm6[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm8, %ymm9
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm8, %ymm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm10
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm11 = xmm10[1],xmm9[1]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm11[0,1,2,3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm11[0,1,2,3,4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],mem[0],ymm4[2],mem[2]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm11
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm12
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm12 = xmm12[0],xmm11[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3],ymm4[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm8[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm6[1],xmm5[1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm11[1],mem[1]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 24(%r9), %ymm5
@@ -238,8 +238,8 @@ define void @store_i64_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm4, 128(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm3, 192(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm6, 64(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm7, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm8, 32(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
@@ -556,11 +556,11 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[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:    vinsertf128 $1, %xmm7, %ymm0, %ymm15
-; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %xmm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],xmm14[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm14
+; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %xmm15
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],xmm15[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm7[0],xmm13[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rcx), %ymm7, %ymm13
@@ -575,36 +575,36 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm13
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm13[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm11 = ymm4[1],ymm11[1],ymm4[3],ymm11[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0,1],ymm11[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm15
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm12, %ymm12
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm15[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm14[0,1],ymm11[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm14
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm12 = xmm12[0],xmm14[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm12, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, %xmm7
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm12, %ymm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm0[2,3],ymm12[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm14[1],xmm9[1]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm15[1],xmm9[1]
 ; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm9
-; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm9 = ymm9[0],ymm14[0],ymm9[2],ymm14[2]
+; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm15
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm9 = ymm9[0],ymm15[0],ymm9[2],ymm15[2]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm0[0,1],ymm9[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 48(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm14 = ymm14[0,0,3,2]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm15 = ymm15[0,0,3,2]
 ; AVX1-ONLY-NEXT:    vmovapd 32(%rax), %ymm1
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm1[2,3],ymm14[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm14[0],ymm0[1],ymm14[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm14[0],mem[0],ymm14[2],mem[2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm1[2,3],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm15[0],ymm0[1],ymm15[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm15
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm15[0],mem[0],ymm15[2],mem[2]
 ; AVX1-ONLY-NEXT:    vmovapd 48(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm6
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm14[2,3]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm14[0],ymm3[0],ymm14[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm15[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm15
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm15[0],ymm3[0],ymm15[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm14[1],xmm5[1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm5
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
@@ -2005,66 +2005,66 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    subq $520, %rsp # imm = 0x208
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm9
+; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm8
+; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vmovaps (%r9), %ymm9
 ; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %ymm10
-; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps 16(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rax), %xmm4
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm5
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm2[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm7
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3],ymm5[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 16(%r8), %xmm2
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm4, %ymm6
+; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 16(%rax), %xmm7
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm4, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3],ymm4[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm6
 ; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1],ymm4[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm9[1],ymm7[1],ymm9[3],ymm7[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm10[1],ymm1[1],ymm10[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm0
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rcx), %ymm1, %ymm3
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm8[1],ymm5[1],ymm8[3],ymm5[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm9[1],ymm2[1],ymm9[3],ymm2[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rcx), %ymm2, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm4
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm12
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm11
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm11[0]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm11[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
 ; 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 32(%rsi), %ymm10
@@ -2078,43 +2078,43 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps 48(%r8), %xmm0
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm9[1],ymm0[1],ymm9[3],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm7
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm7
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm7[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm8[0]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm2
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm1, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm15
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm15[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm13[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm15[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm14[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %ymm5
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm1[1],ymm5[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm0[1],ymm5[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rax), %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %ymm4
 ; AVX1-ONLY-NEXT:    vmovaps 80(%r8), %xmm0
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm4[1],ymm0[1],ymm4[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rcx), %xmm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm14[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm13
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm14
@@ -2134,14 +2134,14 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
 ; 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 {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
@@ -2156,11 +2156,11 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rdx), %xmm9
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm9[0],xmm8[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm9[0],xmm7[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm8[1],xmm6[1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
@@ -2175,8 +2175,8 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovapd 112(%rdi), %xmm6
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm4 = ymm4[0,0,3,2]
-; AVX1-ONLY-NEXT:    vmovapd 96(%rax), %ymm7
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm7[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 96(%rax), %ymm8
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm8[2,3],ymm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0],ymm6[1],ymm4[2],ymm6[3]
 ; AVX1-ONLY-NEXT:    vmovapd 112(%rcx), %xmm6
 ; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm9
@@ -2187,10 +2187,10 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm6[1],mem[1]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 120(%r9), %ymm10
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm10[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0],ymm7[1],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm7[0],xmm15[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0],ymm8[1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm8[0],xmm15[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm10 = xmm10[0],mem[0]
@@ -2202,18 +2202,18 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps %xmm12, 16(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm10, (%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm11, 464(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm7, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 448(%rax)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm4, 832(%rax)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm3, 768(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 576(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm5, 544(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm8, 352(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 352(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm2, 320(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 128(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm14, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm13, 96(%rax)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm6, 864(%rax)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm9, 800(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm13, 736(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm14, 736(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 704(%rax)
 ; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
@@ -4737,76 +4737,77 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm2
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm7
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm6
 ; AVX1-ONLY-NEXT:    vmovaps (%r9), %ymm4
+; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm8
 ; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm9
-; AVX1-ONLY-NEXT:    vmovaps 16(%r8), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 16(%r8), %xmm5
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm10
-; AVX1-ONLY-NEXT:    vmovaps 16(%rax), %xmm8
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm9, %ymm5
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm11
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm11[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm13
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm12, %ymm12
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm12[2,3],ymm5[4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm5[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm12
-; AVX1-ONLY-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm12[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm5[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm5
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm9[0],xmm8[0]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm7, %ymm11
+; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm12
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm7, %ymm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm11[2,3],ymm7[4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm11
+; AVX1-ONLY-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm7[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm7
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm12[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vmovaps 16(%rax), %xmm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm10 = xmm11[1],xmm10[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm12[1]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm8[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:    vunpcklpd {{.*#+}} ymm3 = ymm7[0],ymm4[0],ymm7[2],ymm4[2]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm7
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm10 = xmm10[0],xmm7[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm6[0],ymm4[0],ymm6[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm8
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm8[0],xmm6[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm8[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 32(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm9[1],ymm1[3],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm3
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm10[1],ymm1[3],ymm10[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm4[1],ymm6[1],ymm4[3],ymm6[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm6[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm1
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm3[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm7[0],xmm1[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rcx), %ymm2, %ymm4
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm3
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm0[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm4, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm4
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm0[0],xmm4[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm6, %ymm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm6[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %ymm4
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm5
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %ymm5
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %ymm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm5[0],ymm2[2],ymm5[2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm1
@@ -4816,7 +4817,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; 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 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm4[1],ymm0[1],ymm4[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm0[1],ymm5[3],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rax), %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
@@ -4828,33 +4829,32 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm3
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3],ymm4[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm1
-; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm2
+; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm11
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm11[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
+; 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 64(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %ymm2
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rcx), %xmm4
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm5
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
@@ -4862,12 +4862,12 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 80(%rax), %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 80(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
@@ -4921,32 +4921,32 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %xmm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm12
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps 128(%rax), %xmm3
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3],ymm4[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %ymm1
-; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rcx), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm10
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm10[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 128(%rax), %xmm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %ymm2
+; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rcx), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm14
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm14[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
+; 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 128(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %ymm2
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
 ; AVX1-ONLY-NEXT:    vmovaps 144(%rcx), %xmm4
 ; AVX1-ONLY-NEXT:    vmovaps 144(%rdx), %xmm5
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
@@ -4954,12 +4954,12 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 144(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 144(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 144(%rax), %xmm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 144(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
@@ -5013,39 +5013,39 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 192(%r9), %xmm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm9
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %xmm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm2, %ymm4
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rax), %xmm5
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm4, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3],ymm2[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %ymm4
 ; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm4[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm8
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm8[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm5[1]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %ymm5
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm5[0],ymm1[2],ymm5[2]
+; 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 192(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],mem[0],ymm0[2],mem[2]
 ; AVX1-ONLY-NEXT:    vmovaps 208(%rcx), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 208(%rdx), %xmm6
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm6[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm4[4,5,6,7]
+; 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:    vmovaps 208(%rdi), %xmm4
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm1[1],ymm5[3],ymm1[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 208(%rax), %xmm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm1[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5062,9 +5062,9 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovapd 224(%r8), %ymm2
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rax), %ymm2, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 232(%rcx), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 232(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%r8), %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm0
@@ -5082,26 +5082,26 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vbroadcastsd 216(%r9), %ymm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm15[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovapd 240(%rcx), %xmm15
-; AVX1-ONLY-NEXT:    vmovapd 240(%rdx), %xmm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],xmm15[0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm14[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 240(%r9), %ymm14
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm14[3]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm14 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 240(%rdx), %xmm13
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],xmm15[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm13[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 240(%r9), %ymm13
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm13[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm13 = xmm15[1],mem[1]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 248(%r9), %ymm15
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0],ymm0[1],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm12[0],xmm10[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm10[0],xmm11[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm13[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm12[0],xmm14[0]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm10[0],xmm11[0]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm9[0],xmm8[0]
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm9 = xmm13[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm9 = xmm9[0],mem[0]
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rsi
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm13
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm15
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm12
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm11
@@ -5113,9 +5113,9 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps %xmm11, 1360(%rsi)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm8, 1344(%rsi)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm12, 464(%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %xmm15, 448(%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %xmm13, 912(%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %xmm14, 896(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %xmm14, 448(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, 912(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %xmm13, 896(%rsi)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm0, 1760(%rsi)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm1, 1728(%rsi)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm2, 1696(%rsi)
@@ -10375,38 +10375,38 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride7_vf64:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $3816, %rsp # imm = 0xEE8
+; AVX1-ONLY-NEXT:    subq $3832, %rsp # imm = 0xEF8
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm7
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm6
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %ymm8
-; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps (%r9), %ymm7
+; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm1[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm6
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm2
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm4
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm4, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3],ymm3[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm5
+; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm3, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3],ymm3[4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm4
-; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rax), %xmm2
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm7[1],ymm5[1],ymm7[3],ymm5[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm6[1],ymm4[1],ymm6[3],ymm4[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10414,7 +10414,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm8[1],ymm2[1],ymm8[3],ymm2[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm7[1],ymm2[1],ymm7[3],ymm2[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10459,33 +10459,33 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rax), %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10541,33 +10541,33 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 128(%rax), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 128(%rax), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 144(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 144(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 144(%rax), %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10623,33 +10623,33 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%r9), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 192(%r9), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 192(%rax), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 192(%rax), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 208(%rax), %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10705,33 +10705,33 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 256(%r9), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 256(%r9), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 256(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 256(%rax), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 256(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
+; 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, %ymm1, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 256(%rdx), %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 256(%rax), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdx), %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 264(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 256(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 256(%rsi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 272(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 272(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 272(%rax), %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10787,33 +10787,33 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 320(%r9), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 320(%r9), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 320(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 320(%rax), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 320(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 320(%rdx), %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 320(%rax), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdx), %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 328(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rsi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 336(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 336(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 336(%rax), %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10862,34 +10862,34 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps 368(%r8), %xmm0
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm14[1],ymm0[1],ymm14[3],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 368(%rcx), %xmm13
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 368(%rcx), %xmm12
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm12[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%r9), %xmm12
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm12[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 384(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 384(%rax), %xmm11
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm1, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdx), %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 384(%r9), %xmm13
+; AVX1-ONLY-NEXT:    vmovaps 384(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm13[0]
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 384(%rax), %xmm10
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdx), %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 392(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 384(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%rsi), %ymm10
-; AVX1-ONLY-NEXT:    vmovaps 400(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm10[1],ymm1[1],ymm10[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 384(%rsi), %ymm9
+; AVX1-ONLY-NEXT:    vmovaps 400(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm9[1],ymm0[1],ymm9[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 400(%rax), %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10933,16 +10933,16 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 480(%rsi), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm9
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm11[0],xmm2[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rcx), %ymm1, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 480(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2],ymm3[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 488(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm2[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 488(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%r8), %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -11132,12 +11132,12 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps 352(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm14[0],ymm0[2],ymm14[2]
 ; AVX1-ONLY-NEXT:    vmovaps 368(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm13[0]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[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 384(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm10[0],ymm0[2],ymm10[2]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm12[1],xmm11[1]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm13[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:    vmovaps 384(%r8), %ymm0
@@ -11154,16 +11154,16 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 416(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovapd 432(%rcx), %xmm7
+; AVX1-ONLY-NEXT:    vmovapd 432(%rcx), %xmm6
 ; AVX1-ONLY-NEXT:    vmovapd 432(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm7[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm6[0]
 ; 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 432(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm2[0,0,3,2]
-; AVX1-ONLY-NEXT:    vmovapd 416(%rax), %ymm6
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm6[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 416(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -11178,7 +11178,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovapd 448(%rsi), %ymm2
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm3[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm3[0,1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 464(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm2[0,0,3,2]
@@ -11186,13 +11186,13 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm4[2,3],ymm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2],ymm0[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:    vinsertf128 $1, %xmm11, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1],ymm2[2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovapd 480(%r8), %ymm1
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rax), %ymm1, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 480(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rax), %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 480(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovapd 480(%rsi), %ymm3
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
@@ -11201,42 +11201,42 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovapd 496(%rdi), %xmm5
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm3[0,0,3,2]
-; AVX1-ONLY-NEXT:    vmovapd 480(%rax), %ymm0
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm0[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 480(%rax), %ymm11
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm11[2,3],ymm3[2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2],ymm5[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm7[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 440(%r9), %ymm7
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm6[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 440(%r9), %ymm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 464(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd 464(%rdx), %xmm3
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 464(%r9), %ymm6
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 464(%rcx), %xmm3
-; AVX1-ONLY-NEXT:    vmovapd 464(%rdx), %xmm6
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 464(%r9), %ymm7
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm6[0,1,2],ymm7[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 472(%r9), %ymm7
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm3[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm8[0],ymm4[1],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm7[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 472(%r9), %ymm3
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm9[0],ymm4[1],ymm9[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 496(%rcx), %xmm4
-; AVX1-ONLY-NEXT:    vmovapd 496(%rdx), %xmm7
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],xmm4[0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 496(%r9), %ymm7
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm7[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm4[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 504(%r9), %ymm4
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm1[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0],ymm0[1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 496(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd 496(%rdx), %xmm4
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 496(%r9), %ymm4
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 504(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0],ymm11[1],ymm2[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm11[3]
 ; AVX1-ONLY-NEXT:    vmovupd %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:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
@@ -11245,8 +11245,8 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm4 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm7 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm11 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm3 = xmm0[0],mem[0]
@@ -11264,33 +11264,33 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX1-ONLY-NEXT:    vmovaps 256(%rdx), %xmm14
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm11
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm10
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm12
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm13
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 320(%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 448(%rdx), %xmm9
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdx), %xmm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdx), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdx), %xmm8
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm6
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm6, 16(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, 2704(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, 2704(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, 2688(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm9, 3152(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 3152(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm15, 3136(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, 2256(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm9, 2256(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, 2240(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm13, 1360(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, 1344(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm12, 464(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm7, 448(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm11, 912(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm11, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm10, 912(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm4, 896(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm14, 1808(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm5, 1792(%rax)
@@ -11502,7 +11502,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; 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:    addq $3816, %rsp # imm = 0xEE8
+; AVX1-ONLY-NEXT:    addq $3832, %rsp # imm = 0xEF8
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;

diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-8.ll
index bdf4e30..adcb0f5 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-8.ll
@@ -556,11 +556,11 @@ define void @store_i64_stride8_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r10), %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2],ymm4[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm2
 ; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r10), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
@@ -571,11 +571,11 @@ define void @store_i64_stride8_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdx), %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm2[0],ymm8[1],ymm2[2],ymm8[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm4[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm3, %ymm4
 ; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rdx), %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm3[2,3],ymm4[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm4[2,3],ymm3[4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm6
 ; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm8
@@ -583,8 +583,8 @@ define void @store_i64_stride8_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm4, %ymm9
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r10), %ymm4, %ymm4
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm4 = ymm4[0],ymm9[1],ymm4[2],ymm9[2]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm8[1],xmm6[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm6, %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm8[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r10), %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
@@ -594,8 +594,8 @@ define void @store_i64_stride8_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm6, %ymm10
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, (%rdx), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm10[1],ymm6[2],ymm10[2]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm9[1],xmm8[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm8, %ymm7
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm9[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rdx), %ymm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm7[2,3],ymm8[4,5],ymm7[6,7]
@@ -1392,17 +1392,17 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r10), %ymm9, %ymm9
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm9 = ymm9[0],ymm11[1],ymm9[2],ymm11[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm4, %ymm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm4, %ymm5
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rdx), %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3],ymm4[4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm7, %ymm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm8[1],xmm7[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm4, %ymm5
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r10), %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3],ymm4[4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm4
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm5
@@ -1418,17 +1418,17 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r10), %ymm8, %ymm8
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2],ymm9[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm5[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rdx), %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm6, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm3
 ; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r10), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm3
@@ -1444,18 +1444,18 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rdx), %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm4, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 72(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
@@ -1471,18 +1471,18 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm6, %ymm7
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
-; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 104(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1
@@ -1490,7 +1490,7 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 16(%rcx), %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 16(%r9), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps 16(%r8), %xmm3
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
@@ -1545,18 +1545,18 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm4
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm4[0],xmm1[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm13
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 112(%r9), %xmm14
+; AVX1-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 112(%r9), %xmm15
 ; AVX1-ONLY-NEXT:    vmovaps 112(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm0[0],xmm14[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm0[0],xmm15[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 112(%rax), %ymm12
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1,2,3,4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 120(%rdx), %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm14[1]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm15[1]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 120(%r10), %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rdx
@@ -1571,12 +1571,12 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 992(%rdx)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 960(%rdx)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm12, 928(%rdx)
-; AVX1-ONLY-NEXT:    vmovaps %ymm13, 896(%rdx)
+; AVX1-ONLY-NEXT:    vmovaps %ymm14, 896(%rdx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 864(%rdx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 832(%rdx)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 800(%rdx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 768(%rdx)
@@ -1609,7 +1609,7 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vmovaps %ymm4, 192(%rdx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rdx)
-; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%rdx)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%rdx)
@@ -3040,17 +3040,17 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r10), %ymm7, %ymm7
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm7 = ymm7[0],ymm9[1],ymm7[2],ymm9[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm2, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm2, %ymm3
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rdx), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm4, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm6[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm6[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm2, %ymm3
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r10), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
@@ -3068,17 +3068,17 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r10), %ymm7, %ymm7
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm7 = ymm7[0],ymm9[1],ymm7[2],ymm9[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm4[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm4[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rdx), %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm5, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm6[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm6[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm1, %ymm3
 ; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r10), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3],ymm1[4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm3
@@ -3094,18 +3094,18 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rdx), %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 72(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
@@ -3122,17 +3122,17 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r10), %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 104(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
@@ -3150,17 +3150,17 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%r10), %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 136(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm1
@@ -3178,17 +3178,17 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 168(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 168(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm1
@@ -3206,17 +3206,17 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 200(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm1
@@ -3234,17 +3234,17 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 232(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 232(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1
@@ -6578,73 +6578,73 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, (%r10), %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm4
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm5
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm6
-; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm2
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdx), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm7
-; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm8[0],xmm7[0]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm1, %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm2
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm6[0]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm9
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm3, %ymm10
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r10), %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm3 = ymm3[0],ymm10[1],ymm3[2],ymm10[2]
-; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm3
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm4, %ymm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rdx), %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm7, %ymm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm8[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r10), %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm4
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm5
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm6, %ymm7
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%rdx), %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
-; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm6, %ymm7
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r10), %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
-; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm5[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rdx), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm8, %ymm10
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%r10), %ymm8, %ymm8
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm8 = ymm8[0],ymm10[1],ymm8[2],ymm10[2]
+; AVX1-ONLY-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm8
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm4[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm4
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rdx), %ymm5
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3],ymm3[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm3, %ymm4
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%r10), %ymm5
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3],ymm3[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm5, %ymm6
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%rdx), %ymm5, %ymm5
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm6[1],ymm5[2],ymm6[2]
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm5, %ymm6
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 64(%r10), %ymm5, %ymm5
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm6[1],ymm5[2],ymm6[2]
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm4[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm3, %ymm0
+; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rdx), %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 72(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
@@ -6662,17 +6662,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%r10), %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 104(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
@@ -6690,17 +6690,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 128(%r10), %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 136(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %xmm1
@@ -6718,17 +6718,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 168(%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 168(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm1
@@ -6746,17 +6746,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 192(%r10), %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 200(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm1
@@ -6774,17 +6774,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 232(%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 232(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 256(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm1
@@ -6802,17 +6802,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 256(%r10), %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 264(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 264(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm1
@@ -6830,17 +6830,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%r10), %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 296(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 296(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm1
@@ -6858,17 +6858,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 320(%r10), %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 328(%rdx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 328(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 352(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 352(%rsi), %xmm1
@@ -6886,17 +6886,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 352(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 360(%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 360(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 384(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 384(%rsi), %xmm1
@@ -6914,17 +6914,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 384(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 392(%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 392(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 416(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 416(%rsi), %xmm1
@@ -6942,17 +6942,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 416(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 424(%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 424(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 448(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 448(%rsi), %xmm1
@@ -6970,17 +6970,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 456(%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 456(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 480(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 480(%rsi), %xmm1
@@ -6998,17 +6998,17 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%r10), %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 488(%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm3[1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vbroadcastsd 488(%r10), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1