From 13447d3664f3d91cf18d5bc298656fb8db5172c9 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Wed, 14 Aug 2019 12:36:23 +0000
Subject: [PATCH] [X86] Add missing regular 512-bit vXi8 extract subvector cost
 model tests

These tests don't cover many cases where the subvectors don't start on aligned indices, but that can be added later.

llvm-svn: 368839
---
 .../CostModel/X86/shuffle-extract_subvector.ll     | 494 ++++++++++++++++++---
 1 file changed, 421 insertions(+), 73 deletions(-)

diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-extract_subvector.ll b/llvm/test/Analysis/CostModel/X86/shuffle-extract_subvector.ll
index 9a85cd2..e96e1ad 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-extract_subvector.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-extract_subvector.ll
@@ -619,7 +619,6 @@ define void @test_vXi16(<4 x i16> %src64, <8 x i16> %src128, <16 x i16> %src256,
   ret void
 }
 
-; TODO - add full 512-bit vector test coverage
 define void @test_vXi8(<8 x i8> %src64, <16 x i8> %src128, <32 x i8> %src256, <64 x i8> %src512) {
 ; SSE-LABEL: 'test_vXi8'
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
@@ -692,6 +691,52 @@ define void @test_vXi8(<8 x i8> %src64, <16 x i8> %src128, <32 x i8> %src256, <6
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX-LABEL: 'test_vXi8'
@@ -765,80 +810,291 @@ define void @test_vXi8(<8 x i8> %src64, <16 x i8> %src128, <32 x i8> %src256, <6
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-; AVX512-LABEL: 'test_vXi8'
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; AVX512F-LABEL: 'test_vXi8'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512BW-LABEL: 'test_vXi8'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'test_vXi8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
@@ -911,6 +1167,52 @@ define void @test_vXi8(<8 x i8> %src64, <16 x i8> %src128, <32 x i8> %src256, <6
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
@@ -983,5 +1285,51 @@ define void @test_vXi8(<8 x i8> %src64, <16 x i8> %src128, <32 x i8> %src256, <6
   %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
   %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
   %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
+  %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
+  %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
+  %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
+  %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
+  %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
+  %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
+  %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
+  %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
+  %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
+  %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
+  %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
+  %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
+  %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
+  %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
+  %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
+  %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
+  %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+  %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9,i32 10, i32 11>
+  %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+  %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+  %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+  %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+  %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+  %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+  %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+  %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+  %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+  %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+  %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+  %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+  %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+  %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+  %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+  %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+  %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+  %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+  %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+  %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23,i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+  %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+  %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+  %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23,i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+  %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
   ret void
 }
-- 
2.7.4