[llvm] [X86] combineX86ShuffleChain - always combine to a new VPERMV node if the root shuffle was a VPERMV node (PR #128183)

Simon Pilgrim via llvm-commits llvm-commits at lists.llvm.org
Fri Feb 21 07:01:21 PST 2025


https://github.com/RKSimon created https://github.com/llvm/llvm-project/pull/128183

Similar to what we already do for VPERMV3 nodes - if we're trying to create a new unary variable shuffle and we started with a VPERMV node then always create a new one if it reduces the shuffle chain depth

>From 9bda6b094f263e7a82f4fa586314ada4aeedba4a Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev at redking.me.uk>
Date: Fri, 21 Feb 2025 15:00:01 +0000
Subject: [PATCH] [X86] combineX86ShuffleChain - always combine to a new VPERMV
 node if the root shuffle was a VPERMV node

Similar to what we already do for VPERMV3 nodes - if we're trying to create a new unary variable shuffle and we started with a VPERMV node then always create a new one if it reduces the shuffle chain depth
---
 llvm/lib/Target/X86/X86ISelLowering.cpp       |   5 +-
 .../vector-interleaved-store-i16-stride-7.ll  |  57 ++++---
 .../vector-interleaved-store-i32-stride-7.ll  |  61 ++++----
 .../vector-interleaved-store-i8-stride-5.ll   | 142 +++++++++---------
 .../vector-interleaved-store-i8-stride-7.ll   | 115 +++++++-------
 .../vector-interleaved-store-i8-stride-8.ll   |  84 ++++++-----
 6 files changed, 231 insertions(+), 233 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index dd666f15bf57f..f8f7d3e367896 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -40090,8 +40090,9 @@ static SDValue combineX86ShuffleChain(ArrayRef<SDValue> Inputs, SDValue Root,
   bool AllowBWIVPERMV3 =
       (Depth >= BWIVPERMV3ShuffleDepth || HasSlowVariableMask);
 
-  // If root was a VPERMV3 node, always allow a variable shuffle.
-  if (Root.getOpcode() == X86ISD::VPERMV3)
+  // If root was a VPERMV/VPERMV3 node, always allow a variable shuffle.
+  if ((UnaryShuffle && Root.getOpcode() == X86ISD::VPERMV) ||
+      Root.getOpcode() == X86ISD::VPERMV3)
     AllowVariableCrossLaneMask = AllowVariablePerLaneMask = true;
 
   bool MaskContainsZeros = isAnyZero(Mask);
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
index 351d98540c2a5..3f8f061f359f9 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
@@ -588,20 +588,19 @@ define void @store_i16_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX2-FCP-NEXT:    vmovq {{.*#+}} xmm5 = mem[0],zero
 ; AVX2-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
-; AVX2-FCP-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [5,7,1,3,7,0,0,0]
-; AVX2-FCP-NEXT:    vpermd %ymm4, %ymm1, %ymm1
-; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,1,4,5],zero,zero,zero,zero,zero,zero,ymm1[10,11,14,15,2,3,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,3,5,7,1,3,5,7]
-; AVX2-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
-; AVX2-FCP-NEXT:    vpermd %ymm0, %ymm5, %ymm5
-; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = zero,zero,zero,zero,ymm5[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm5[18,19,22,23,26,27],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX2-FCP-NEXT:    vpor %ymm5, %ymm1, %ymm1
+; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [1,5,3,7,1,5,3,7]
+; AVX2-FCP-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX2-FCP-NEXT:    vpermd %ymm0, %ymm1, %ymm1
+; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,ymm1[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[18,19,22,23,26,27],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [5,7,1,3,7,0,0,0]
+; AVX2-FCP-NEXT:    vpermd %ymm4, %ymm5, %ymm5
+; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[0,1,4,5],zero,zero,zero,zero,zero,zero,ymm5[10,11,14,15,2,3,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-FCP-NEXT:    vpor %ymm1, %ymm5, %ymm1
 ; AVX2-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm3, %ymm2
 ; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,ymm2[0,1,8,9,4,5,6,7,4,5],zero,zero,ymm2[26,27],zero,zero,zero,zero,ymm2[24,25,20,21,22,23,20,21,28,29]
 ; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm4[0,1,8,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm4[2,3],zero,zero,ymm4[18,19,26,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX2-FCP-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,2,4,6,0,2,4,6]
+; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,4,2,6,0,4,2,6]
 ; AVX2-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
 ; AVX2-FCP-NEXT:    vpermd %ymm0, %ymm3, %ymm0
 ; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,0,1,4,5,8,9,u,u,u,u,u,u,u,u,18,19,22,23,26,27,u,u,u,u]
@@ -670,17 +669,16 @@ define void @store_i16_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm4
 ; AVX512-FCP-NEXT:    vmovq {{.*#+}} xmm5 = mem[0],zero
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
-; AVX512-FCP-NEXT:    vpermi2q %ymm5, %ymm0, %ymm1
-; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [0,2,4,6,0,2,4,6]
-; AVX512-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
-; AVX512-FCP-NEXT:    vpermd %ymm1, %ymm0, %ymm0
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[18,19,22,23,26,27],zero,zero,zero,zero
-; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,3,5,7,1,3,5,7]
+; AVX512-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
+; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,4,2,6,0,4,2,6]
+; AVX512-FCP-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX512-FCP-NEXT:    vpermd %ymm0, %ymm1, %ymm1
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[18,19,22,23,26,27],zero,zero,zero,zero
+; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,5,3,7,1,5,3,7]
 ; AVX512-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
-; AVX512-FCP-NEXT:    vpermd %ymm1, %ymm5, %ymm1
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,ymm1[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[18,19,22,23,26,27],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512-FCP-NEXT:    vpermd %ymm0, %ymm5, %ymm0
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,ymm0[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[18,19,22,23,26,27],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm4[0,1,8,9],zero,zero,zero,zero,ymm4[u,u,u,u,u,u,2,3],zero,zero,ymm4[18,19,26,27,u,u,u,u,u,u],zero,zero,zero,zero
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm3, %ymm2
 ; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,ymm2[0,1,8,9,u,u,u,u,u,u],zero,zero,ymm2[26,27],zero,zero,zero,zero,ymm2[u,u,u,u,u,u,20,21,28,29]
@@ -753,17 +751,16 @@ define void @store_i16_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm4
 ; AVX512DQ-FCP-NEXT:    vmovq {{.*#+}} xmm5 = mem[0],zero
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
-; AVX512DQ-FCP-NEXT:    vpermi2q %ymm5, %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [0,2,4,6,0,2,4,6]
-; AVX512DQ-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm1, %ymm0, %ymm0
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[18,19,22,23,26,27],zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,3,5,7,1,3,5,7]
+; AVX512DQ-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
+; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,4,2,6,0,4,2,6]
+; AVX512DQ-FCP-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm0, %ymm1, %ymm1
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[18,19,22,23,26,27],zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,5,3,7,1,5,3,7]
 ; AVX512DQ-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm1, %ymm5, %ymm1
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,ymm1[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[18,19,22,23,26,27],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512DQ-FCP-NEXT:    vpermd %ymm0, %ymm5, %ymm0
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,ymm0[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[18,19,22,23,26,27],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm4[0,1,8,9],zero,zero,zero,zero,ymm4[u,u,u,u,u,u,2,3],zero,zero,ymm4[18,19,26,27,u,u,u,u,u,u],zero,zero,zero,zero
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm3, %ymm2
 ; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,ymm2[0,1,8,9,u,u,u,u,u,u],zero,zero,ymm2[26,27],zero,zero,zero,zero,ymm2[u,u,u,u,u,u,20,21,28,29]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
index 6476c3139daa7..34100adacbeb9 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
@@ -95,17 +95,17 @@ define void @store_i32_stride7_vf2(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX2-NEXT:    vmovsd {{.*#+}} xmm3 = mem[0],zero
 ; AVX2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2,2,1,4,6,6,5]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,1]
-; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,2]
-; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6],ymm1[7]
-; AVX2-NEXT:    vmovaps {{.*#+}} xmm3 = [3,5,7,u]
-; AVX2-NEXT:    vpermps %ymm2, %ymm3, %ymm2
-; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [3,5,0,1,3,5,0,1]
+; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [0,4,2,6,0,4,2,6]
+; AVX2-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX2-NEXT:    vpermps %ymm0, %ymm1, %ymm1
+; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[0,2,2,1,4,6,6,5]
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,2,1]
+; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6],ymm3[7]
+; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [5,3,0,1,5,3,0,1]
 ; AVX2-NEXT:    # ymm3 = mem[0,1,0,1]
 ; AVX2-NEXT:    vpermps %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vmovaps {{.*#+}} xmm3 = [3,5,7,u]
+; AVX2-NEXT:    vpermps %ymm2, %ymm3, %ymm2
 ; AVX2-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm0[3]
 ; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-NEXT:    vmovlps %xmm0, 48(%rax)
@@ -130,17 +130,17 @@ define void @store_i32_stride7_vf2(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX2-FP-NEXT:    vmovsd {{.*#+}} xmm3 = mem[0],zero
 ; AVX2-FP-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
-; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2,2,1,4,6,6,5]
-; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,1]
-; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm3 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,2]
-; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6],ymm1[7]
-; AVX2-FP-NEXT:    vmovaps {{.*#+}} xmm3 = [3,5,7,u]
-; AVX2-FP-NEXT:    vpermps %ymm2, %ymm3, %ymm2
-; AVX2-FP-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [3,5,0,1,3,5,0,1]
+; AVX2-FP-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [0,4,2,6,0,4,2,6]
+; AVX2-FP-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX2-FP-NEXT:    vpermps %ymm0, %ymm1, %ymm1
+; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[0,2,2,1,4,6,6,5]
+; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,2,1]
+; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6],ymm3[7]
+; AVX2-FP-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [5,3,0,1,5,3,0,1]
 ; AVX2-FP-NEXT:    # ymm3 = mem[0,1,0,1]
 ; AVX2-FP-NEXT:    vpermps %ymm0, %ymm3, %ymm0
+; AVX2-FP-NEXT:    vmovaps {{.*#+}} xmm3 = [3,5,7,u]
+; AVX2-FP-NEXT:    vpermps %ymm2, %ymm3, %ymm2
 ; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm0[3]
 ; AVX2-FP-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-FP-NEXT:    vmovlps %xmm0, 48(%rax)
@@ -165,23 +165,22 @@ define void @store_i32_stride7_vf2(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX2-FCP-NEXT:    vmovsd {{.*#+}} xmm3 = mem[0],zero
 ; AVX2-FCP-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
-; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm1 = [3,5,7,u]
-; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm1, %ymm1
-; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [3,5,0,1,3,5,0,1]
-; AVX2-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
-; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm3, %ymm3
-; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm3[3]
-; AVX2-FCP-NEXT:    vmovaps {{.*#+}} ymm4 = [0,2,4,6,u,u,u,1]
-; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm4, %ymm2
-; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm4 = [0,2,4,0,0,2,4,0]
+; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [5,3,0,1,5,3,0,1]
+; AVX2-FCP-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm1, %ymm1
+; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm3 = [3,5,7,u]
+; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm3, %ymm3
+; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1,2],xmm1[3]
+; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm4 = [0,4,2,0,0,4,2,0]
 ; AVX2-FCP-NEXT:    # ymm4 = mem[0,1,0,1]
 ; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm4, %ymm0
+; AVX2-FCP-NEXT:    vmovaps {{.*#+}} ymm4 = [0,2,4,6,u,u,u,1]
+; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm4, %ymm2
 ; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6],ymm2[7]
-; AVX2-FCP-NEXT:    vextractf128 $1, %ymm3, %xmm2
-; AVX2-FCP-NEXT:    vmovlps %xmm2, 48(%rax)
+; AVX2-FCP-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX2-FCP-NEXT:    vmovlps %xmm1, 48(%rax)
 ; AVX2-FCP-NEXT:    vmovaps %ymm0, (%rax)
-; AVX2-FCP-NEXT:    vmovaps %xmm1, 32(%rax)
+; AVX2-FCP-NEXT:    vmovaps %xmm3, 32(%rax)
 ; AVX2-FCP-NEXT:    vzeroupper
 ; AVX2-FCP-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll
index f7a44fea5b02b..01a2f7f46f939 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll
@@ -5576,8 +5576,8 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512BW-FCP-NEXT:    vpshufb %xmm14, %xmm2, %xmm3
 ; AVX512BW-FCP-NEXT:    vmovdqa (%rdx), %xmm12
 ; AVX512BW-FCP-NEXT:    vmovdqa 32(%rdx), %xmm4
-; AVX512BW-FCP-NEXT:    vmovdqa64 {{.*#+}} xmm17 = [6,128,8,128,u,7,128,9,128,11,128,u,10,128,12,128]
-; AVX512BW-FCP-NEXT:    vpshufb %xmm17, %xmm4, %xmm5
+; AVX512BW-FCP-NEXT:    vmovdqa64 {{.*#+}} xmm16 = [6,128,8,128,u,7,128,9,128,11,128,u,10,128,12,128]
+; AVX512BW-FCP-NEXT:    vpshufb %xmm16, %xmm4, %xmm5
 ; AVX512BW-FCP-NEXT:    vpor %xmm3, %xmm5, %xmm3
 ; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,1,1]
 ; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm3, %zmm1
@@ -5585,15 +5585,15 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512BW-FCP-NEXT:    vmovdqa 32(%rdi), %xmm3
 ; AVX512BW-FCP-NEXT:    vmovdqa64 {{.*#+}} xmm19 = [8,128,u,7,128,9,128,u,128,u,10,128,12,128,u,11]
 ; AVX512BW-FCP-NEXT:    vpshufb %xmm19, %xmm3, %xmm9
-; AVX512BW-FCP-NEXT:    vmovdqa64 (%rsi), %xmm18
+; AVX512BW-FCP-NEXT:    vmovdqa64 (%rsi), %xmm17
 ; AVX512BW-FCP-NEXT:    vmovdqa 32(%rsi), %xmm5
 ; AVX512BW-FCP-NEXT:    vmovdqa64 {{.*#+}} xmm20 = [128,8,u,128,7,128,9,u,11,u,128,10,128,12,u,128]
 ; AVX512BW-FCP-NEXT:    vpshufb %xmm20, %xmm5, %xmm10
 ; AVX512BW-FCP-NEXT:    vpor %xmm9, %xmm10, %xmm9
 ; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} ymm15 = ymm9[0,0,1,1]
-; AVX512BW-FCP-NEXT:    vmovdqa64 32(%rdi), %ymm16
+; AVX512BW-FCP-NEXT:    vmovdqa64 32(%rdi), %ymm18
 ; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm9 = [3,3,3,0,4,4,4,4]
-; AVX512BW-FCP-NEXT:    vpermd %ymm16, %ymm9, %ymm22
+; AVX512BW-FCP-NEXT:    vpermd %ymm18, %ymm9, %ymm22
 ; AVX512BW-FCP-NEXT:    vmovdqa64 32(%rsi), %ymm23
 ; AVX512BW-FCP-NEXT:    vpbroadcastq {{.*#+}} ymm10 = [0,0,13,2,15,0,1,14,0,0,13,2,15,0,1,14,0,0,13,2,15,0,1,14,0,0,13,2,15,0,1,14]
 ; AVX512BW-FCP-NEXT:    movl $138547332, %eax # imm = 0x8421084
@@ -5611,9 +5611,9 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm22 = zmm23[0,1,2,3],mem[4,5,6,7]
 ; AVX512BW-FCP-NEXT:    vmovdqa64 {{.*#+}} zmm15 = [u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,19,128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,27,128,128,26,128,28,128,128,128,128,29,128,31,128,128,30]
 ; AVX512BW-FCP-NEXT:    vpshufb %zmm15, %zmm22, %zmm22
-; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm16[0,1,2,3],mem[4,5,6,7]
-; AVX512BW-FCP-NEXT:    vmovdqa64 {{.*#+}} zmm16 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128]
-; AVX512BW-FCP-NEXT:    vpshufb %zmm16, %zmm23, %zmm23
+; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm18[0,1,2,3],mem[4,5,6,7]
+; AVX512BW-FCP-NEXT:    vmovdqa64 {{.*#+}} zmm18 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128]
+; AVX512BW-FCP-NEXT:    vpshufb %zmm18, %zmm23, %zmm23
 ; AVX512BW-FCP-NEXT:    vporq %zmm22, %zmm23, %zmm22
 ; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} zmm22 = zmm22[2,2,3,3,6,6,7,7]
 ; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],mem[4,5,6,7]
@@ -5633,66 +5633,65 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512BW-FCP-NEXT:    kmovq %rax, %k4
 ; AVX512BW-FCP-NEXT:    vmovdqu8 %zmm21, %zmm8 {%k4}
 ; AVX512BW-FCP-NEXT:    vpshufb %xmm14, %xmm11, %xmm14
-; AVX512BW-FCP-NEXT:    vpshufb %xmm17, %xmm12, %xmm17
-; AVX512BW-FCP-NEXT:    vporq %xmm14, %xmm17, %xmm14
+; AVX512BW-FCP-NEXT:    vpshufb %xmm16, %xmm12, %xmm16
+; AVX512BW-FCP-NEXT:    vporq %xmm14, %xmm16, %xmm14
 ; AVX512BW-FCP-NEXT:    vpunpcklbw {{.*#+}} xmm11 = xmm11[0],xmm12[0],xmm11[1],xmm12[1],xmm11[2],xmm12[2],xmm11[3],xmm12[3],xmm11[4],xmm12[4],xmm11[5],xmm12[5],xmm11[6],xmm12[6],xmm11[7],xmm12[7]
 ; AVX512BW-FCP-NEXT:    vmovdqa {{.*#+}} xmm12 = [2,u,1,0,5,4,u,3,u,7,6,11,10,u,9,8]
 ; AVX512BW-FCP-NEXT:    vpshufb %xmm12, %xmm11, %xmm11
 ; AVX512BW-FCP-NEXT:    vinserti32x4 $2, %xmm14, %zmm11, %zmm11
 ; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} zmm14 = zmm11[0,0,1,1,4,4,5,5]
 ; AVX512BW-FCP-NEXT:    vpshufb %xmm19, %xmm13, %xmm11
-; AVX512BW-FCP-NEXT:    vpshufb %xmm20, %xmm18, %xmm17
-; AVX512BW-FCP-NEXT:    vporq %xmm11, %xmm17, %xmm11
-; AVX512BW-FCP-NEXT:    vpunpcklbw {{.*#+}} xmm13 = xmm13[0],xmm18[0],xmm13[1],xmm18[1],xmm13[2],xmm18[2],xmm13[3],xmm18[3],xmm13[4],xmm18[4],xmm13[5],xmm18[5],xmm13[6],xmm18[6],xmm13[7],xmm18[7]
-; AVX512BW-FCP-NEXT:    vmovdqa64 {{.*#+}} xmm17 = [0,1,4,5,u,2,3,6,7,10,11,u,8,9,12,13]
-; AVX512BW-FCP-NEXT:    vpshufb %xmm17, %xmm13, %xmm13
+; AVX512BW-FCP-NEXT:    vpshufb %xmm20, %xmm17, %xmm16
+; AVX512BW-FCP-NEXT:    vporq %xmm11, %xmm16, %xmm11
+; AVX512BW-FCP-NEXT:    vpunpcklbw {{.*#+}} xmm13 = xmm13[0],xmm17[0],xmm13[1],xmm17[1],xmm13[2],xmm17[2],xmm13[3],xmm17[3],xmm13[4],xmm17[4],xmm13[5],xmm17[5],xmm13[6],xmm17[6],xmm13[7],xmm17[7]
+; AVX512BW-FCP-NEXT:    vmovdqa64 {{.*#+}} xmm16 = [0,1,4,5,u,2,3,6,7,10,11,u,8,9,12,13]
+; AVX512BW-FCP-NEXT:    vpshufb %xmm16, %xmm13, %xmm13
 ; AVX512BW-FCP-NEXT:    vinserti32x4 $2, %xmm11, %zmm13, %zmm11
 ; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} zmm11 = zmm11[0,0,1,1,4,4,5,5]
 ; AVX512BW-FCP-NEXT:    movabsq $-4165393823095705204, %rax # imm = 0xC6318C6318C6318C
 ; AVX512BW-FCP-NEXT:    kmovq %rax, %k4
 ; AVX512BW-FCP-NEXT:    vmovdqu8 %zmm14, %zmm11 {%k4}
-; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm13 = mem[0,1,2,3,0,1,2,3]
-; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} zmm14 = [0,0,0,0,0,0,1,1,9,9,10,10,10,10,10,10]
-; AVX512BW-FCP-NEXT:    vpermd %zmm13, %zmm14, %zmm14
+; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} zmm13 = [0,0,0,0,0,0,1,1,1,1,2,2,2,2,2,2]
+; AVX512BW-FCP-NEXT:    vpermd %zmm0, %zmm13, %zmm13
 ; AVX512BW-FCP-NEXT:    movabsq $595056260442243600, %rax # imm = 0x842108421084210
 ; AVX512BW-FCP-NEXT:    kmovq %rax, %k4
-; AVX512BW-FCP-NEXT:    vmovdqu8 %zmm14, %zmm11 {%k4}
-; AVX512BW-FCP-NEXT:    vmovdqa (%rdx), %ymm14
-; AVX512BW-FCP-NEXT:    vpshufb %ymm6, %ymm14, %ymm6
-; AVX512BW-FCP-NEXT:    vmovdqa64 (%rcx), %ymm18
-; AVX512BW-FCP-NEXT:    vpshufb %ymm7, %ymm18, %ymm7
+; AVX512BW-FCP-NEXT:    vmovdqu8 %zmm13, %zmm11 {%k4}
+; AVX512BW-FCP-NEXT:    vmovdqa (%rdx), %ymm13
+; AVX512BW-FCP-NEXT:    vpshufb %ymm6, %ymm13, %ymm6
+; AVX512BW-FCP-NEXT:    vmovdqa (%rcx), %ymm14
+; AVX512BW-FCP-NEXT:    vpshufb %ymm7, %ymm14, %ymm7
 ; AVX512BW-FCP-NEXT:    vpor %ymm6, %ymm7, %ymm6
-; AVX512BW-FCP-NEXT:    vpshufb %ymm23, %ymm14, %ymm7
-; AVX512BW-FCP-NEXT:    vpshufb %ymm24, %ymm18, %ymm19
-; AVX512BW-FCP-NEXT:    vporq %ymm7, %ymm19, %ymm7
+; AVX512BW-FCP-NEXT:    vpshufb %ymm23, %ymm13, %ymm7
+; AVX512BW-FCP-NEXT:    vpshufb %ymm24, %ymm14, %ymm17
+; AVX512BW-FCP-NEXT:    vporq %ymm7, %ymm17, %ymm7
 ; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[2,2,3,3]
 ; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm6, %zmm6
 ; AVX512BW-FCP-NEXT:    vmovdqa (%rsi), %ymm7
 ; AVX512BW-FCP-NEXT:    vpshufb %ymm15, %ymm7, %ymm15
-; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %ymm19
-; AVX512BW-FCP-NEXT:    vpshufb %ymm16, %ymm19, %ymm16
-; AVX512BW-FCP-NEXT:    vporq %ymm15, %ymm16, %ymm15
+; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %ymm17
+; AVX512BW-FCP-NEXT:    vpshufb %ymm18, %ymm17, %ymm18
+; AVX512BW-FCP-NEXT:    vporq %ymm15, %ymm18, %ymm15
 ; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} ymm15 = ymm15[2,2,3,3]
-; AVX512BW-FCP-NEXT:    vpermd %ymm19, %ymm9, %ymm9
+; AVX512BW-FCP-NEXT:    vpermd %ymm17, %ymm9, %ymm9
 ; AVX512BW-FCP-NEXT:    vpshufb %ymm10, %ymm7, %ymm9 {%k1}
 ; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm15, %zmm9, %zmm9
 ; AVX512BW-FCP-NEXT:    vmovdqu8 %zmm6, %zmm9 {%k2}
-; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} zmm6 = [3,3,3,3,0,4,4,4,12,14,13,13,13,13,12,14]
-; AVX512BW-FCP-NEXT:    vpermd %zmm13, %zmm6, %zmm6
+; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} zmm6 = [3,3,3,3,0,4,4,4,4,6,5,5,5,5,4,6]
+; AVX512BW-FCP-NEXT:    vpermd %zmm0, %zmm6, %zmm6
 ; AVX512BW-FCP-NEXT:    movabsq $1190112520884487201, %rax # imm = 0x1084210842108421
 ; AVX512BW-FCP-NEXT:    kmovq %rax, %k1
 ; AVX512BW-FCP-NEXT:    vmovdqu8 %zmm6, %zmm9 {%k1}
 ; AVX512BW-FCP-NEXT:    vpunpcklbw {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3],xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
 ; AVX512BW-FCP-NEXT:    vpshufb %xmm12, %xmm2, %xmm2
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = ymm18[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,25],zero,ymm18[27],zero,zero,ymm18[26],zero,ymm18[28],zero,ymm18[30],zero,zero,ymm18[29],zero,ymm18[31],zero
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm14[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,ymm14[27],zero,zero,ymm14[26],zero,ymm14[28],zero,ymm14[30],zero,zero,ymm14[29],zero,ymm14[31],zero,zero
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = ymm14[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,25],zero,ymm14[27],zero,zero,ymm14[26],zero,ymm14[28],zero,ymm14[30],zero,zero,ymm14[29],zero,ymm14[31],zero
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,ymm13[27],zero,zero,ymm13[26],zero,ymm13[28],zero,ymm13[30],zero,zero,ymm13[29],zero,ymm13[31],zero,zero
 ; AVX512BW-FCP-NEXT:    vpor %ymm4, %ymm6, %ymm4
 ; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm6 = [2,2,3,3,8,8,9,9]
 ; AVX512BW-FCP-NEXT:    vpermt2q %zmm2, %zmm6, %zmm4
 ; AVX512BW-FCP-NEXT:    vpunpcklbw {{.*#+}} xmm2 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3],xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
-; AVX512BW-FCP-NEXT:    vpshufb %xmm17, %xmm2, %xmm2
+; AVX512BW-FCP-NEXT:    vpshufb %xmm16, %xmm2, %xmm2
 ; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,27],zero,zero,ymm7[26],zero,ymm7[28],zero,zero,zero,zero,ymm7[29],zero,ymm7[31],zero,zero,ymm7[30]
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm19[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,zero,ymm19[26],zero,ymm19[28],zero,zero,ymm19[27],zero,ymm19[29],zero,ymm19[31],zero,zero,ymm19[30],zero
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm17[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,zero,ymm17[26],zero,ymm17[28],zero,zero,ymm17[27],zero,ymm17[29],zero,ymm17[31],zero,zero,ymm17[30],zero
 ; AVX512BW-FCP-NEXT:    vpor %ymm3, %ymm5, %ymm3
 ; AVX512BW-FCP-NEXT:    vpermt2q %zmm2, %zmm6, %zmm3
 ; AVX512BW-FCP-NEXT:    vmovdqu8 %zmm4, %zmm3 {%k3}
@@ -5899,8 +5898,8 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm14, %xmm2, %xmm3
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%rdx), %xmm12
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa 32(%rdx), %xmm4
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 {{.*#+}} xmm17 = [6,128,8,128,u,7,128,9,128,11,128,u,10,128,12,128]
-; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm17, %xmm4, %xmm5
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 {{.*#+}} xmm16 = [6,128,8,128,u,7,128,9,128,11,128,u,10,128,12,128]
+; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm16, %xmm4, %xmm5
 ; AVX512DQ-BW-FCP-NEXT:    vpor %xmm3, %xmm5, %xmm3
 ; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,1,1]
 ; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm3, %zmm1
@@ -5908,15 +5907,15 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa 32(%rdi), %xmm3
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 {{.*#+}} xmm19 = [8,128,u,7,128,9,128,u,128,u,10,128,12,128,u,11]
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm19, %xmm3, %xmm9
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rsi), %xmm18
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rsi), %xmm17
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa 32(%rsi), %xmm5
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 {{.*#+}} xmm20 = [128,8,u,128,7,128,9,u,11,u,128,10,128,12,u,128]
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm20, %xmm5, %xmm10
 ; AVX512DQ-BW-FCP-NEXT:    vpor %xmm9, %xmm10, %xmm9
 ; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} ymm15 = ymm9[0,0,1,1]
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 32(%rdi), %ymm16
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 32(%rdi), %ymm18
 ; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm9 = [3,3,3,0,4,4,4,4]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm16, %ymm9, %ymm22
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm18, %ymm9, %ymm22
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 32(%rsi), %ymm23
 ; AVX512DQ-BW-FCP-NEXT:    vpbroadcastq {{.*#+}} ymm10 = [0,0,13,2,15,0,1,14,0,0,13,2,15,0,1,14,0,0,13,2,15,0,1,14,0,0,13,2,15,0,1,14]
 ; AVX512DQ-BW-FCP-NEXT:    movl $138547332, %eax # imm = 0x8421084
@@ -5934,9 +5933,9 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm22 = zmm23[0,1,2,3],mem[4,5,6,7]
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 {{.*#+}} zmm15 = [u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,19,128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,27,128,128,26,128,28,128,128,128,128,29,128,31,128,128,30]
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb %zmm15, %zmm22, %zmm22
-; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm16[0,1,2,3],mem[4,5,6,7]
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 {{.*#+}} zmm16 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128]
-; AVX512DQ-BW-FCP-NEXT:    vpshufb %zmm16, %zmm23, %zmm23
+; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm18[0,1,2,3],mem[4,5,6,7]
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 {{.*#+}} zmm18 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128]
+; AVX512DQ-BW-FCP-NEXT:    vpshufb %zmm18, %zmm23, %zmm23
 ; AVX512DQ-BW-FCP-NEXT:    vporq %zmm22, %zmm23, %zmm22
 ; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} zmm22 = zmm22[2,2,3,3,6,6,7,7]
 ; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],mem[4,5,6,7]
@@ -5956,66 +5955,65 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512DQ-BW-FCP-NEXT:    kmovq %rax, %k4
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqu8 %zmm21, %zmm8 {%k4}
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm14, %xmm11, %xmm14
-; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm17, %xmm12, %xmm17
-; AVX512DQ-BW-FCP-NEXT:    vporq %xmm14, %xmm17, %xmm14
+; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm16, %xmm12, %xmm16
+; AVX512DQ-BW-FCP-NEXT:    vporq %xmm14, %xmm16, %xmm14
 ; AVX512DQ-BW-FCP-NEXT:    vpunpcklbw {{.*#+}} xmm11 = xmm11[0],xmm12[0],xmm11[1],xmm12[1],xmm11[2],xmm12[2],xmm11[3],xmm12[3],xmm11[4],xmm12[4],xmm11[5],xmm12[5],xmm11[6],xmm12[6],xmm11[7],xmm12[7]
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa {{.*#+}} xmm12 = [2,u,1,0,5,4,u,3,u,7,6,11,10,u,9,8]
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm12, %xmm11, %xmm11
 ; AVX512DQ-BW-FCP-NEXT:    vinserti32x4 $2, %xmm14, %zmm11, %zmm11
 ; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} zmm14 = zmm11[0,0,1,1,4,4,5,5]
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm19, %xmm13, %xmm11
-; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm20, %xmm18, %xmm17
-; AVX512DQ-BW-FCP-NEXT:    vporq %xmm11, %xmm17, %xmm11
-; AVX512DQ-BW-FCP-NEXT:    vpunpcklbw {{.*#+}} xmm13 = xmm13[0],xmm18[0],xmm13[1],xmm18[1],xmm13[2],xmm18[2],xmm13[3],xmm18[3],xmm13[4],xmm18[4],xmm13[5],xmm18[5],xmm13[6],xmm18[6],xmm13[7],xmm18[7]
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 {{.*#+}} xmm17 = [0,1,4,5,u,2,3,6,7,10,11,u,8,9,12,13]
-; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm17, %xmm13, %xmm13
+; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm20, %xmm17, %xmm16
+; AVX512DQ-BW-FCP-NEXT:    vporq %xmm11, %xmm16, %xmm11
+; AVX512DQ-BW-FCP-NEXT:    vpunpcklbw {{.*#+}} xmm13 = xmm13[0],xmm17[0],xmm13[1],xmm17[1],xmm13[2],xmm17[2],xmm13[3],xmm17[3],xmm13[4],xmm17[4],xmm13[5],xmm17[5],xmm13[6],xmm17[6],xmm13[7],xmm17[7]
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 {{.*#+}} xmm16 = [0,1,4,5,u,2,3,6,7,10,11,u,8,9,12,13]
+; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm16, %xmm13, %xmm13
 ; AVX512DQ-BW-FCP-NEXT:    vinserti32x4 $2, %xmm11, %zmm13, %zmm11
 ; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} zmm11 = zmm11[0,0,1,1,4,4,5,5]
 ; AVX512DQ-BW-FCP-NEXT:    movabsq $-4165393823095705204, %rax # imm = 0xC6318C6318C6318C
 ; AVX512DQ-BW-FCP-NEXT:    kmovq %rax, %k4
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqu8 %zmm14, %zmm11 {%k4}
-; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm13 = mem[0,1,2,3,0,1,2,3]
-; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} zmm14 = [0,0,0,0,0,0,1,1,9,9,10,10,10,10,10,10]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %zmm13, %zmm14, %zmm14
+; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} zmm13 = [0,0,0,0,0,0,1,1,1,1,2,2,2,2,2,2]
+; AVX512DQ-BW-FCP-NEXT:    vpermd %zmm0, %zmm13, %zmm13
 ; AVX512DQ-BW-FCP-NEXT:    movabsq $595056260442243600, %rax # imm = 0x842108421084210
 ; AVX512DQ-BW-FCP-NEXT:    kmovq %rax, %k4
-; AVX512DQ-BW-FCP-NEXT:    vmovdqu8 %zmm14, %zmm11 {%k4}
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%rdx), %ymm14
-; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm6, %ymm14, %ymm6
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rcx), %ymm18
-; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm7, %ymm18, %ymm7
+; AVX512DQ-BW-FCP-NEXT:    vmovdqu8 %zmm13, %zmm11 {%k4}
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%rdx), %ymm13
+; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm6, %ymm13, %ymm6
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%rcx), %ymm14
+; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm7, %ymm14, %ymm7
 ; AVX512DQ-BW-FCP-NEXT:    vpor %ymm6, %ymm7, %ymm6
-; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm23, %ymm14, %ymm7
-; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm24, %ymm18, %ymm19
-; AVX512DQ-BW-FCP-NEXT:    vporq %ymm7, %ymm19, %ymm7
+; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm23, %ymm13, %ymm7
+; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm24, %ymm14, %ymm17
+; AVX512DQ-BW-FCP-NEXT:    vporq %ymm7, %ymm17, %ymm7
 ; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[2,2,3,3]
 ; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm6, %zmm6
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%rsi), %ymm7
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm15, %ymm7, %ymm15
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %ymm19
-; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm16, %ymm19, %ymm16
-; AVX512DQ-BW-FCP-NEXT:    vporq %ymm15, %ymm16, %ymm15
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %ymm17
+; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm18, %ymm17, %ymm18
+; AVX512DQ-BW-FCP-NEXT:    vporq %ymm15, %ymm18, %ymm15
 ; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} ymm15 = ymm15[2,2,3,3]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm19, %ymm9, %ymm9
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm17, %ymm9, %ymm9
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm10, %ymm7, %ymm9 {%k1}
 ; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm15, %zmm9, %zmm9
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqu8 %zmm6, %zmm9 {%k2}
-; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} zmm6 = [3,3,3,3,0,4,4,4,12,14,13,13,13,13,12,14]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %zmm13, %zmm6, %zmm6
+; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} zmm6 = [3,3,3,3,0,4,4,4,4,6,5,5,5,5,4,6]
+; AVX512DQ-BW-FCP-NEXT:    vpermd %zmm0, %zmm6, %zmm6
 ; AVX512DQ-BW-FCP-NEXT:    movabsq $1190112520884487201, %rax # imm = 0x1084210842108421
 ; AVX512DQ-BW-FCP-NEXT:    kmovq %rax, %k1
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqu8 %zmm6, %zmm9 {%k1}
 ; AVX512DQ-BW-FCP-NEXT:    vpunpcklbw {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3],xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm12, %xmm2, %xmm2
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = ymm18[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,25],zero,ymm18[27],zero,zero,ymm18[26],zero,ymm18[28],zero,ymm18[30],zero,zero,ymm18[29],zero,ymm18[31],zero
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm14[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,ymm14[27],zero,zero,ymm14[26],zero,ymm14[28],zero,ymm14[30],zero,zero,ymm14[29],zero,ymm14[31],zero,zero
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = ymm14[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,25],zero,ymm14[27],zero,zero,ymm14[26],zero,ymm14[28],zero,ymm14[30],zero,zero,ymm14[29],zero,ymm14[31],zero
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,ymm13[27],zero,zero,ymm13[26],zero,ymm13[28],zero,ymm13[30],zero,zero,ymm13[29],zero,ymm13[31],zero,zero
 ; AVX512DQ-BW-FCP-NEXT:    vpor %ymm4, %ymm6, %ymm4
 ; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm6 = [2,2,3,3,8,8,9,9]
 ; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm2, %zmm6, %zmm4
 ; AVX512DQ-BW-FCP-NEXT:    vpunpcklbw {{.*#+}} xmm2 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3],xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
-; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm17, %xmm2, %xmm2
+; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm16, %xmm2, %xmm2
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,27],zero,zero,ymm7[26],zero,ymm7[28],zero,zero,zero,zero,ymm7[29],zero,ymm7[31],zero,zero,ymm7[30]
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm19[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,zero,ymm19[26],zero,ymm19[28],zero,zero,ymm19[27],zero,ymm19[29],zero,ymm19[31],zero,zero,ymm19[30],zero
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm17[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,zero,ymm17[26],zero,ymm17[28],zero,zero,ymm17[27],zero,ymm17[29],zero,ymm17[31],zero,zero,ymm17[30],zero
 ; AVX512DQ-BW-FCP-NEXT:    vpor %ymm3, %ymm5, %ymm3
 ; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm2, %zmm6, %zmm3
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqu8 %zmm4, %zmm3 {%k3}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
index 89ed0040a71c2..c1dba071b4353 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
@@ -1000,20 +1000,19 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; AVX2-FCP-NEXT:    vmovq {{.*#+}} xmm5 = mem[0],zero
 ; AVX2-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
-; AVX2-FCP-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [1,3,5,7,1,3,5,7]
-; AVX2-FCP-NEXT:    # ymm1 = mem[0,1,0,1]
-; AVX2-FCP-NEXT:    vpermd %ymm4, %ymm1, %ymm1
-; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,ymm1[1,5,9,13],zero,zero,zero,ymm1[2,6,10,14],zero,zero,zero,ymm1[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX2-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,3,5,0,5,1,3,0]
-; AVX2-FCP-NEXT:    vpermd %ymm0, %ymm5, %ymm5
-; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[0,4,8],zero,zero,zero,zero,ymm5[1,5,9],zero,zero,zero,zero,ymm5[2,6,18],zero,zero,zero,zero,ymm5[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX2-FCP-NEXT:    vpor %ymm5, %ymm1, %ymm1
+; AVX2-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [1,5,3,0,3,1,5,0]
+; AVX2-FCP-NEXT:    vpermd %ymm0, %ymm1, %ymm1
+; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4,8],zero,zero,zero,zero,ymm1[1,5,9],zero,zero,zero,zero,ymm1[2,6,18],zero,zero,zero,zero,ymm1[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,3,5,7,1,3,5,7]
+; AVX2-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
+; AVX2-FCP-NEXT:    vpermd %ymm4, %ymm5, %ymm5
+; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = zero,zero,zero,ymm5[1,5,9,13],zero,zero,zero,ymm5[2,6,10,14],zero,zero,zero,ymm5[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-FCP-NEXT:    vpor %ymm1, %ymm5, %ymm1
 ; AVX2-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm3, %ymm2
 ; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28],zero,zero
 ; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm4[0,8],zero,zero,zero,zero,zero,ymm4[1,9],zero,zero,zero,zero,zero,ymm4[2,10,18,26],zero,zero,zero,zero,zero,ymm4[19,27],zero,zero,zero,zero,zero,ymm4[20,28]
 ; AVX2-FCP-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,2,4,6,0,2,4,6]
+; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,4,2,6,0,4,2,6]
 ; AVX2-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
 ; AVX2-FCP-NEXT:    vpermd %ymm0, %ymm3, %ymm0
 ; AVX2-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
@@ -1082,16 +1081,15 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm4
 ; AVX512-FCP-NEXT:    vmovq {{.*#+}} xmm5 = mem[0],zero
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
-; AVX512-FCP-NEXT:    vpermi2q %ymm5, %ymm0, %ymm1
-; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [0,2,4,6,0,2,4,6]
-; AVX512-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
-; AVX512-FCP-NEXT:    vpermd %ymm1, %ymm0, %ymm0
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,ymm0[0,4,8],zero,zero,zero,zero,ymm0[1,5,9],zero,zero,zero,zero,ymm0[18,22,26],zero,zero,zero,zero,ymm0[19,23,27],zero,zero,zero,zero
-; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,3,5,0,5,1,3,0]
-; AVX512-FCP-NEXT:    vpermd %ymm1, %ymm5, %ymm1
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4,8],zero,zero,zero,zero,ymm1[1,5,9],zero,zero,zero,zero,ymm1[2,6,18],zero,zero,zero,zero,ymm1[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
+; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,4,2,6,0,4,2,6]
+; AVX512-FCP-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX512-FCP-NEXT:    vpermd %ymm0, %ymm1, %ymm1
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,ymm1[0,4,8],zero,zero,zero,zero,ymm1[1,5,9],zero,zero,zero,zero,ymm1[18,22,26],zero,zero,zero,zero,ymm1[19,23,27],zero,zero,zero,zero
+; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,5,3,0,3,1,5,0]
+; AVX512-FCP-NEXT:    vpermd %ymm0, %ymm5, %ymm0
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,4,8],zero,zero,zero,zero,ymm0[1,5,9],zero,zero,zero,zero,ymm0[2,6,18],zero,zero,zero,zero,ymm0[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm4[0,8],zero,zero,ymm4[u,u,u,1,9],zero,zero,ymm4[u,u,u,2,10,18,26,u,u,u],zero,zero,ymm4[19,27,u,u,u],zero,zero,ymm4[20,28]
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm3, %ymm2
 ; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,ymm2[0,8,u,u,u],zero,zero,ymm2[1,9,u,u,u],zero,zero,zero,zero,ymm2[u,u,u,19,27],zero,zero,ymm2[u,u,u,20,28],zero,zero
@@ -1165,16 +1163,15 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm4
 ; AVX512DQ-FCP-NEXT:    vmovq {{.*#+}} xmm5 = mem[0],zero
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
-; AVX512DQ-FCP-NEXT:    vpermi2q %ymm5, %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [0,2,4,6,0,2,4,6]
-; AVX512DQ-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm1, %ymm0, %ymm0
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,ymm0[0,4,8],zero,zero,zero,zero,ymm0[1,5,9],zero,zero,zero,zero,ymm0[18,22,26],zero,zero,zero,zero,ymm0[19,23,27],zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,3,5,0,5,1,3,0]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm1, %ymm5, %ymm1
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4,8],zero,zero,zero,zero,ymm1[1,5,9],zero,zero,zero,zero,ymm1[2,6,18],zero,zero,zero,zero,ymm1[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512DQ-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
+; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,4,2,6,0,4,2,6]
+; AVX512DQ-FCP-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm0, %ymm1, %ymm1
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,ymm1[0,4,8],zero,zero,zero,zero,ymm1[1,5,9],zero,zero,zero,zero,ymm1[18,22,26],zero,zero,zero,zero,ymm1[19,23,27],zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,5,3,0,3,1,5,0]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm0, %ymm5, %ymm0
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,4,8],zero,zero,zero,zero,ymm0[1,5,9],zero,zero,zero,zero,ymm0[2,6,18],zero,zero,zero,zero,ymm0[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm4[0,8],zero,zero,ymm4[u,u,u,1,9],zero,zero,ymm4[u,u,u,2,10,18,26,u,u,u],zero,zero,ymm4[19,27,u,u,u],zero,zero,ymm4[20,28]
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm3, %ymm2
 ; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,ymm2[0,8,u,u,u],zero,zero,ymm2[1,9,u,u,u],zero,zero,zero,zero,ymm2[u,u,u,19,27],zero,zero,ymm2[u,u,u,20,28],zero,zero
@@ -1244,30 +1241,29 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; AVX512BW-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm4
 ; AVX512BW-FCP-NEXT:    vmovq {{.*#+}} xmm5 = mem[0],zero
 ; AVX512BW-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
-; AVX512BW-FCP-NEXT:    vpermi2q %ymm5, %ymm0, %ymm1
-; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [1,3,5,7,1,3,5,7]
-; AVX512BW-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
-; AVX512BW-FCP-NEXT:    vpermd %ymm4, %ymm0, %ymm0
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,ymm0[1,5,9,13],zero,zero,zero,ymm0[2,6,10,14],zero,zero,zero,ymm0[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,3,5,0,5,1,3,0]
-; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm5, %ymm5
+; AVX512BW-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
+; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [1,3,5,7,1,3,5,7]
+; AVX512BW-FCP-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX512BW-FCP-NEXT:    vpermd %ymm4, %ymm1, %ymm1
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,ymm1[1,5,9,13],zero,zero,zero,ymm1[2,6,10,14],zero,zero,zero,ymm1[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,5,3,0,3,1,5,0]
+; AVX512BW-FCP-NEXT:    vpermd %ymm0, %ymm5, %ymm5
 ; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[0,4,8],zero,zero,zero,zero,ymm5[1,5,9],zero,zero,zero,zero,ymm5[2,6,18],zero,zero,zero,zero,ymm5[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-FCP-NEXT:    vpor %ymm0, %ymm5, %ymm0
+; AVX512BW-FCP-NEXT:    vpor %ymm1, %ymm5, %ymm1
 ; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[0,8],zero,zero,zero,zero,zero,ymm4[1,9],zero,zero,zero,zero,zero,ymm4[2,10,18,26],zero,zero,zero,zero,zero,ymm4[19,27],zero,zero,zero,zero,zero,ymm4[20,28]
 ; AVX512BW-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm3, %ymm2
 ; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28],zero,zero
 ; AVX512BW-FCP-NEXT:    vpor %ymm4, %ymm2, %ymm2
-; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,2,4,6,0,2,4,6]
+; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,4,2,6,0,4,2,6]
 ; AVX512BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
-; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
+; AVX512BW-FCP-NEXT:    vpermd %ymm0, %ymm3, %ymm0
 ; AVX512BW-FCP-NEXT:    movl $236730480, %ecx # imm = 0xE1C3870
 ; AVX512BW-FCP-NEXT:    kmovd %ecx, %k1
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 {%k1} = ymm1[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 {%k1} = ymm0[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
 ; AVX512BW-FCP-NEXT:    vmovdqa %ymm2, (%rax)
-; AVX512BW-FCP-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX512BW-FCP-NEXT:    vmovq %xmm1, 48(%rax)
-; AVX512BW-FCP-NEXT:    vmovdqa %xmm0, 32(%rax)
+; AVX512BW-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm0
+; AVX512BW-FCP-NEXT:    vmovq %xmm0, 48(%rax)
+; AVX512BW-FCP-NEXT:    vmovdqa %xmm1, 32(%rax)
 ; AVX512BW-FCP-NEXT:    vzeroupper
 ; AVX512BW-FCP-NEXT:    retq
 ;
@@ -1323,30 +1319,29 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm4
 ; AVX512DQ-BW-FCP-NEXT:    vmovq {{.*#+}} xmm5 = mem[0],zero
 ; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
-; AVX512DQ-BW-FCP-NEXT:    vpermi2q %ymm5, %ymm0, %ymm1
-; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [1,3,5,7,1,3,5,7]
-; AVX512DQ-BW-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm4, %ymm0, %ymm0
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,ymm0[1,5,9,13],zero,zero,zero,ymm0[2,6,10,14],zero,zero,zero,ymm0[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,3,5,0,5,1,3,0]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm5, %ymm5
+; AVX512DQ-BW-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
+; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [1,3,5,7,1,3,5,7]
+; AVX512DQ-BW-FCP-NEXT:    # ymm1 = mem[0,1,0,1]
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm4, %ymm1, %ymm1
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = zero,zero,zero,ymm1[1,5,9,13],zero,zero,zero,ymm1[2,6,10,14],zero,zero,zero,ymm1[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,5,3,0,3,1,5,0]
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm0, %ymm5, %ymm5
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[0,4,8],zero,zero,zero,zero,ymm5[1,5,9],zero,zero,zero,zero,ymm5[2,6,18],zero,zero,zero,zero,ymm5[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-BW-FCP-NEXT:    vpor %ymm0, %ymm5, %ymm0
+; AVX512DQ-BW-FCP-NEXT:    vpor %ymm1, %ymm5, %ymm1
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[0,8],zero,zero,zero,zero,zero,ymm4[1,9],zero,zero,zero,zero,zero,ymm4[2,10,18,26],zero,zero,zero,zero,zero,ymm4[19,27],zero,zero,zero,zero,zero,ymm4[20,28]
 ; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm3, %ymm2
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28],zero,zero
 ; AVX512DQ-BW-FCP-NEXT:    vpor %ymm4, %ymm2, %ymm2
-; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,2,4,6,0,2,4,6]
+; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,4,2,6,0,4,2,6]
 ; AVX512DQ-BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm0, %ymm3, %ymm0
 ; AVX512DQ-BW-FCP-NEXT:    movl $236730480, %ecx # imm = 0xE1C3870
 ; AVX512DQ-BW-FCP-NEXT:    kmovd %ecx, %k1
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 {%k1} = ymm1[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 {%k1} = ymm0[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm2, (%rax)
-; AVX512DQ-BW-FCP-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm1, 48(%rax)
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm0, 32(%rax)
+; AVX512DQ-BW-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm0
+; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm0, 48(%rax)
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm1, 32(%rax)
 ; AVX512DQ-BW-FCP-NEXT:    vzeroupper
 ; AVX512DQ-BW-FCP-NEXT:    retq
   %in.vec0 = load <8 x i8>, ptr %in.vecptr0, align 64
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-8.ll
index 75c470a6d40c6..bb236b375778c 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-8.ll
@@ -2122,44 +2122,48 @@ define void @store_i8_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512BW-FCP-NEXT:    vmovdqa (%r11), %xmm3
 ; AVX512BW-FCP-NEXT:    vinserti128 $1, (%rcx), %ymm1, %ymm1
 ; AVX512BW-FCP-NEXT:    vinserti128 $1, (%rsi), %ymm0, %ymm0
-; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm4
+; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm1
 ; AVX512BW-FCP-NEXT:    vinserti128 $1, (%r10), %ymm3, %ymm3
 ; AVX512BW-FCP-NEXT:    vinserti128 $1, (%r9), %ymm2, %ymm2
-; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm2, %zmm5
+; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm2, %zmm3
+; AVX512BW-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm4 = [4,6,4,6,4,6,4,6]
+; AVX512BW-FCP-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-FCP-NEXT:    vpermq %zmm3, %zmm4, %zmm5
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} zmm5 = zmm5[u,u,u,u,u,u,0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63]
 ; AVX512BW-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm6 = [0,2,0,2,0,2,0,2]
 ; AVX512BW-FCP-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
-; AVX512BW-FCP-NEXT:    vpermq %zmm5, %zmm6, %zmm5
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} zmm5 = zmm5[u,u,u,u,0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u]
 ; AVX512BW-FCP-NEXT:    vpermq %zmm3, %zmm6, %zmm7
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} zmm7 = zmm7[u,u,u,u,u,u,0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63]
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} zmm7 = zmm7[u,u,u,u,0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u]
 ; AVX512BW-FCP-NEXT:    movl $-2004318072, %ecx # imm = 0x88888888
 ; AVX512BW-FCP-NEXT:    kmovd %ecx, %k1
-; AVX512BW-FCP-NEXT:    vmovdqu16 %zmm7, %zmm5 {%k1}
-; AVX512BW-FCP-NEXT:    vpermq %zmm4, %zmm6, %zmm4
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} zmm4 = zmm4[0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u,u,u,u,u]
-; AVX512BW-FCP-NEXT:    vpermq %zmm1, %zmm6, %zmm6
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} zmm6 = zmm6[u,u,0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u,u,u]
+; AVX512BW-FCP-NEXT:    vmovdqu16 %zmm5, %zmm7 {%k1}
+; AVX512BW-FCP-NEXT:    vpermq %zmm1, %zmm4, %zmm4
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} zmm4 = zmm4[u,u,0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u,u,u]
+; AVX512BW-FCP-NEXT:    vpermq %zmm1, %zmm6, %zmm5
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} zmm5 = zmm5[0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u,u,u,u,u]
 ; AVX512BW-FCP-NEXT:    movl $572662306, %ecx # imm = 0x22222222
 ; AVX512BW-FCP-NEXT:    kmovd %ecx, %k2
-; AVX512BW-FCP-NEXT:    vmovdqu16 %zmm6, %zmm4 {%k2}
+; AVX512BW-FCP-NEXT:    vmovdqu16 %zmm4, %zmm5 {%k2}
 ; AVX512BW-FCP-NEXT:    movw $-21846, %cx # imm = 0xAAAA
 ; AVX512BW-FCP-NEXT:    kmovd %ecx, %k3
-; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm5, %zmm4 {%k3}
-; AVX512BW-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm5 = [1,3,1,3,1,3,1,3]
-; AVX512BW-FCP-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
-; AVX512BW-FCP-NEXT:    vpermq %zmm3, %zmm5, %zmm3
+; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm7, %zmm5 {%k3}
+; AVX512BW-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm4 = [5,7,5,7,5,7,5,7]
+; AVX512BW-FCP-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-FCP-NEXT:    vpermq %zmm3, %zmm4, %zmm3
 ; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} zmm3 = zmm3[u,u,u,u,0,8,0,8,u,u,u,u,1,9,1,9,u,u,u,u,18,26,18,26,u,u,u,u,19,27,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63]
-; AVX512BW-FCP-NEXT:    vpermq %zmm2, %zmm5, %zmm2
+; AVX512BW-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm6 = [1,3,1,3,1,3,1,3]
+; AVX512BW-FCP-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512BW-FCP-NEXT:    vpermq %zmm2, %zmm6, %zmm2
 ; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} zmm2 = zmm2[u,u,u,u,0,8,0,8,u,u,u,u,1,9,1,9,u,u,u,u,18,26,18,26,u,u,u,u,19,27,19,27,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u]
 ; AVX512BW-FCP-NEXT:    vmovdqu16 %zmm3, %zmm2 {%k1}
-; AVX512BW-FCP-NEXT:    vpermq %zmm1, %zmm5, %zmm1
+; AVX512BW-FCP-NEXT:    vpermq %zmm1, %zmm4, %zmm1
 ; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} zmm1 = zmm1[0,8,0,8,u,u,u,u,1,9,1,9,u,u,u,u,18,26,18,26,u,u,u,u,19,27,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u,u,u]
-; AVX512BW-FCP-NEXT:    vpermq %zmm0, %zmm5, %zmm0
+; AVX512BW-FCP-NEXT:    vpermq %zmm0, %zmm6, %zmm0
 ; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,8,0,8,u,u,u,u,1,9,1,9,u,u,u,u,18,26,18,26,u,u,u,u,19,27,19,27,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u,u,u,u,u]
 ; AVX512BW-FCP-NEXT:    vmovdqu16 %zmm1, %zmm0 {%k2}
 ; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm2, %zmm0 {%k3}
 ; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, 64(%rax)
-; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, (%rax)
+; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, (%rax)
 ; AVX512BW-FCP-NEXT:    vzeroupper
 ; AVX512BW-FCP-NEXT:    retq
 ;
@@ -2226,44 +2230,48 @@ define void @store_i8_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%r11), %xmm3
 ; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, (%rcx), %ymm1, %ymm1
 ; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, (%rsi), %ymm0, %ymm0
-; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm4
+; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm1
 ; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, (%r10), %ymm3, %ymm3
 ; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, (%r9), %ymm2, %ymm2
-; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm2, %zmm5
+; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm2, %zmm3
+; AVX512DQ-BW-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm4 = [4,6,4,6,4,6,4,6]
+; AVX512DQ-BW-FCP-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm3, %zmm4, %zmm5
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} zmm5 = zmm5[u,u,u,u,u,u,0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63]
 ; AVX512DQ-BW-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm6 = [0,2,0,2,0,2,0,2]
 ; AVX512DQ-BW-FCP-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
-; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm5, %zmm6, %zmm5
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} zmm5 = zmm5[u,u,u,u,0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u]
 ; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm3, %zmm6, %zmm7
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} zmm7 = zmm7[u,u,u,u,u,u,0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63]
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} zmm7 = zmm7[u,u,u,u,0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u]
 ; AVX512DQ-BW-FCP-NEXT:    movl $-2004318072, %ecx # imm = 0x88888888
 ; AVX512DQ-BW-FCP-NEXT:    kmovd %ecx, %k1
-; AVX512DQ-BW-FCP-NEXT:    vmovdqu16 %zmm7, %zmm5 {%k1}
-; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm4, %zmm6, %zmm4
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} zmm4 = zmm4[0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u,u,u,u,u]
-; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm1, %zmm6, %zmm6
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} zmm6 = zmm6[u,u,0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u,u,u]
+; AVX512DQ-BW-FCP-NEXT:    vmovdqu16 %zmm5, %zmm7 {%k1}
+; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm1, %zmm4, %zmm4
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} zmm4 = zmm4[u,u,0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u,u,u]
+; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm1, %zmm6, %zmm5
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} zmm5 = zmm5[0,8,u,u,u,u,u,u,1,9,u,u,u,u,u,u,18,26,u,u,u,u,u,u,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u,u,u,u,u]
 ; AVX512DQ-BW-FCP-NEXT:    movl $572662306, %ecx # imm = 0x22222222
 ; AVX512DQ-BW-FCP-NEXT:    kmovd %ecx, %k2
-; AVX512DQ-BW-FCP-NEXT:    vmovdqu16 %zmm6, %zmm4 {%k2}
+; AVX512DQ-BW-FCP-NEXT:    vmovdqu16 %zmm4, %zmm5 {%k2}
 ; AVX512DQ-BW-FCP-NEXT:    movw $-21846, %cx # imm = 0xAAAA
 ; AVX512DQ-BW-FCP-NEXT:    kmovd %ecx, %k3
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm5, %zmm4 {%k3}
-; AVX512DQ-BW-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm5 = [1,3,1,3,1,3,1,3]
-; AVX512DQ-BW-FCP-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
-; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm3, %zmm5, %zmm3
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm7, %zmm5 {%k3}
+; AVX512DQ-BW-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm4 = [5,7,5,7,5,7,5,7]
+; AVX512DQ-BW-FCP-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm3, %zmm4, %zmm3
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} zmm3 = zmm3[u,u,u,u,0,8,0,8,u,u,u,u,1,9,1,9,u,u,u,u,18,26,18,26,u,u,u,u,19,27,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63]
-; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm2, %zmm5, %zmm2
+; AVX512DQ-BW-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm6 = [1,3,1,3,1,3,1,3]
+; AVX512DQ-BW-FCP-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm2, %zmm6, %zmm2
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} zmm2 = zmm2[u,u,u,u,0,8,0,8,u,u,u,u,1,9,1,9,u,u,u,u,18,26,18,26,u,u,u,u,19,27,19,27,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u]
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqu16 %zmm3, %zmm2 {%k1}
-; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm1, %zmm5, %zmm1
+; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm1, %zmm4, %zmm1
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} zmm1 = zmm1[0,8,0,8,u,u,u,u,1,9,1,9,u,u,u,u,18,26,18,26,u,u,u,u,19,27,19,27,u,u,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u,u,u]
-; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm0, %zmm5, %zmm0
+; AVX512DQ-BW-FCP-NEXT:    vpermq %zmm0, %zmm6, %zmm0
 ; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,8,0,8,u,u,u,u,1,9,1,9,u,u,u,u,18,26,18,26,u,u,u,u,19,27,19,27,u,u,u,u,36,44,u,u,u,u,u,u,37,45,u,u,u,u,u,u,54,62,u,u,u,u,u,u,55,63,u,u,u,u,u,u]
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqu16 %zmm1, %zmm0 {%k2}
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm2, %zmm0 {%k3}
 ; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, 64(%rax)
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, (%rax)
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, (%rax)
 ; AVX512DQ-BW-FCP-NEXT:    vzeroupper
 ; AVX512DQ-BW-FCP-NEXT:    retq
   %in.vec0 = load <16 x i8>, ptr %in.vecptr0, align 64



More information about the llvm-commits mailing list