From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Fri, 5 Oct 2018 14:41:00 +0000 (+0000)
Subject: [X86][AVX] getFauxShuffleMask - add support for INSERT_SUBVECTOR subvector shuffles
X-Git-Tag: android-x86-9.0-r1~12169
X-Git-Url: http://git.osdn.net/view?a=commitdiff_plain;h=d327e07926eecfa668e036dcd290da33ffb68856;p=android-x86%2Fexternal-llvm.git

[X86][AVX] getFauxShuffleMask - add support for INSERT_SUBVECTOR subvector shuffles

Decode subvector shuffles from INSERT_SUBVECTOR(SRC0, SHUFFLE(EXTRACT_SUBVECTOR(SRC1))

This was found necessary while investigating PR39161

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@343853 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 45f87bbaada..91fa216c053 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -6380,6 +6380,42 @@ static bool getFauxShuffleMask(SDValue N, SmallVectorImpl<int> &Mask,
       Ops.push_back(Op);
     return true;
   }
+  case ISD::INSERT_SUBVECTOR: {
+    // Handle INSERT_SUBVECTOR(SRC0, SHUFFLE(EXTRACT_SUBVECTOR(SRC1)) where
+    // SRC0/SRC1 are both of the same valuetype VT.
+    // TODO - add peekThroughOneUseBitcasts support.
+    SDValue Src = N.getOperand(0);
+    SDValue Sub = N.getOperand(1);
+    EVT SubVT = Sub.getValueType();
+    unsigned NumSubElts = SubVT.getVectorNumElements();
+    if (!isa<ConstantSDNode>(N.getOperand(2)) ||
+        !N->isOnlyUserOf(Sub.getNode()))
+      return false;
+    SmallVector<int, 64> SubMask;
+    SmallVector<SDValue, 2> SubInputs;
+    if (!resolveTargetShuffleInputs(Sub, SubInputs, SubMask, DAG))
+      return false;
+    Ops.push_back(Src);
+    for (SDValue &SubInput : SubInputs) {
+      if (SubInput.getOpcode() != ISD::EXTRACT_SUBVECTOR ||
+          SubInput.getOperand(0).getValueType() != VT ||
+          !isa<ConstantSDNode>(SubInput.getOperand(1)))
+        return false;
+      Ops.push_back(SubInput.getOperand(0));
+    }
+    int InsertIdx = N.getConstantOperandVal(2);
+    for (int i = 0; i != (int)NumElts; ++i)
+      Mask.push_back(i);
+    for (int i = 0; i != (int)NumSubElts; ++i) {
+      int M = SubMask[i];
+      if (0 <= M) {
+        int InputIdx = M / NumSubElts;
+        int ExtractIdx = SubInputs[InputIdx].getConstantOperandVal(1);
+        Mask[i + InsertIdx] = (NumElts * (1 + InputIdx)) + ExtractIdx + M;
+      }
+    }
+    return true;
+  }
   case ISD::SCALAR_TO_VECTOR: {
     // Match against a scalar_to_vector of an extract from a vector,
     // for PEXTRW/PEXTRB we must handle the implicit zext of the scalar.
diff --git a/test/CodeGen/X86/avx512-shuffles/partial_permute.ll b/test/CodeGen/X86/avx512-shuffles/partial_permute.ll
index 56319dc55cb..d198fe87bed 100644
--- a/test/CodeGen/X86/avx512-shuffles/partial_permute.ll
+++ b/test/CodeGen/X86/avx512-shuffles/partial_permute.ll
@@ -3138,8 +3138,7 @@ define <8 x float> @test_masked_16xfloat_to_8xfloat_perm_mask2(<16 x float> %vec
 ; CHECK-LABEL: test_masked_16xfloat_to_8xfloat_perm_mask2:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vextractf64x4 $1, %zmm0, %ymm3
-; CHECK-NEXT:    vmovddup {{.*#+}} xmm3 = xmm3[0,0]
-; CHECK-NEXT:    vmovaps {{.*#+}} ymm4 = [0,4,10,11,6,1,4,4]
+; CHECK-NEXT:    vmovaps {{.*#+}} ymm4 = [0,4,8,9,6,1,4,4]
 ; CHECK-NEXT:    vpermi2ps %ymm3, %ymm0, %ymm4
 ; CHECK-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; CHECK-NEXT:    vcmpeqps %ymm0, %ymm2, %k1
@@ -3154,9 +3153,8 @@ define <8 x float> @test_masked_16xfloat_to_8xfloat_perm_mask2(<16 x float> %vec
 define <8 x float> @test_masked_z_16xfloat_to_8xfloat_perm_mask2(<16 x float> %vec, <8 x float> %mask) {
 ; CHECK-LABEL: test_masked_z_16xfloat_to_8xfloat_perm_mask2:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vextractf64x4 $1, %zmm0, %ymm2
-; CHECK-NEXT:    vmovddup {{.*#+}} xmm3 = xmm2[0,0]
-; CHECK-NEXT:    vmovaps {{.*#+}} ymm2 = [0,4,10,11,6,1,4,4]
+; CHECK-NEXT:    vextractf64x4 $1, %zmm0, %ymm3
+; CHECK-NEXT:    vmovaps {{.*#+}} ymm2 = [0,4,8,9,6,1,4,4]
 ; CHECK-NEXT:    vxorps %xmm4, %xmm4, %xmm4
 ; CHECK-NEXT:    vcmpeqps %ymm4, %ymm1, %k1
 ; CHECK-NEXT:    vpermi2ps %ymm3, %ymm0, %ymm2 {%k1} {z}
@@ -3459,10 +3457,9 @@ define <8 x float> @test_masked_16xfloat_to_8xfloat_perm_mem_mask2(<16 x float>*
 ; CHECK-LABEL: test_masked_16xfloat_to_8xfloat_perm_mem_mask2:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps (%rdi), %zmm2
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm3 = xmm2[1,0,0,3]
-; CHECK-NEXT:    vextractf64x4 $1, %zmm2, %ymm2
-; CHECK-NEXT:    vmovaps {{.*#+}} ymm4 = [8,5,2,3,2,9,10,1]
-; CHECK-NEXT:    vpermi2ps %ymm3, %ymm2, %ymm4
+; CHECK-NEXT:    vextractf64x4 $1, %zmm2, %ymm3
+; CHECK-NEXT:    vmovaps {{.*#+}} ymm4 = [9,5,2,3,2,8,8,1]
+; CHECK-NEXT:    vpermi2ps %ymm2, %ymm3, %ymm4
 ; CHECK-NEXT:    vxorps %xmm2, %xmm2, %xmm2
 ; CHECK-NEXT:    vcmpeqps %ymm2, %ymm1, %k1
 ; CHECK-NEXT:    vmovaps %ymm4, %ymm0 {%k1}
@@ -3477,10 +3474,9 @@ define <8 x float> @test_masked_16xfloat_to_8xfloat_perm_mem_mask2(<16 x float>*
 define <8 x float> @test_masked_z_16xfloat_to_8xfloat_perm_mem_mask2(<16 x float>* %vp, <8 x float> %mask) {
 ; CHECK-LABEL: test_masked_z_16xfloat_to_8xfloat_perm_mem_mask2:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vmovaps (%rdi), %zmm1
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[1,0,0,3]
-; CHECK-NEXT:    vextractf64x4 $1, %zmm1, %ymm3
-; CHECK-NEXT:    vmovaps {{.*#+}} ymm1 = [8,5,2,3,2,9,10,1]
+; CHECK-NEXT:    vmovaps (%rdi), %zmm2
+; CHECK-NEXT:    vextractf64x4 $1, %zmm2, %ymm3
+; CHECK-NEXT:    vmovaps {{.*#+}} ymm1 = [9,5,2,3,2,8,8,1]
 ; CHECK-NEXT:    vxorps %xmm4, %xmm4, %xmm4
 ; CHECK-NEXT:    vcmpeqps %ymm4, %ymm0, %k1
 ; CHECK-NEXT:    vpermi2ps %ymm2, %ymm3, %ymm1 {%k1} {z}
diff --git a/test/CodeGen/X86/oddshuffles.ll b/test/CodeGen/X86/oddshuffles.ll
index def822aac5e..1e12ed9778a 100644
--- a/test/CodeGen/X86/oddshuffles.ll
+++ b/test/CodeGen/X86/oddshuffles.ll
@@ -1648,8 +1648,8 @@ define void @interleave_24i32_in(<24 x i32>* %p, <8 x i32>* %q1, <8 x i32>* %q2,
 ; AVX2-FAST-NEXT:    vmovups (%rsi), %ymm0
 ; AVX2-FAST-NEXT:    vmovups (%rdx), %ymm1
 ; AVX2-FAST-NEXT:    vmovups (%rcx), %ymm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm1[1,0,2,2]
-; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,1]
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm3 = [1,0,2,2,1,0,2,2]
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm3, %ymm3
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm0[0,0,2,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm2, %ymm4
diff --git a/test/CodeGen/X86/vector-shuffle-256-v32.ll b/test/CodeGen/X86/vector-shuffle-256-v32.ll
index 159335f6325..3e36b4a3b6a 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v32.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v32.ll
@@ -32,11 +32,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -50,11 +68,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -68,11 +104,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_03_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,3,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_03_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,3,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_03_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,3,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_03_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,3,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_03_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,3,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 3, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -86,11 +140,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_04_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,4,0,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_04_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,4,0,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_04_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,4,0,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_04_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,4,0,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_04_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,4,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 4, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -104,11 +176,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_05_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_05_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_05_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_05_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_05_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 5, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -122,11 +212,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_06_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_06_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_06_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_06_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_06_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 6, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -140,11 +248,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -158,11 +284,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,8]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,8]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,8]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,8]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 8, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -176,11 +320,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_09_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,9,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_09_00_00_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,9,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_09_00_00_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,9,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_09_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,9,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_09_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,9,0,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 9, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -194,11 +356,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,10,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,10,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,10,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,10,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,10,0,0,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 10, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -212,11 +392,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,11,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,11,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,11,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,11,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,11,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 11, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -230,11 +428,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_12_00_00_00_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,12,0,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_12_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,12,0,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_12_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,12,0,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_12_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,12,0,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_12_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,12,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 12, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -248,11 +464,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_13_00_00_00_00_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,13,0,0,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_13_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,13,0,0,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_13_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,13,0,0,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_13_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,13,0,0,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_13_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,13,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 13, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -266,11 +500,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,14,0,0,0,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,14,0,0,0,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,14,0,0,0,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,14,0,0,0,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 14, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -286,11 +538,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,15,0,0,0,0,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,15,0,0,0,0,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,15,0,0,0,0,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,15,0,0,0,0,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,15,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -1118,11 +1388,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -1134,11 +1422,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,2,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -1150,11 +1456,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -1166,11 +1490,29 @@ define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 8, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 8, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -1182,11 +1524,29 @@ define <32 x i8> @shuffle_v32i8_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,14,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 14, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 14, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
@@ -1200,13 +1560,35 @@ define <32 x i8> @shuffle_v32i8_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    movl $15, %eax
-; AVX2OR512VL-NEXT:    vmovd %eax, %xmm1
-; AVX2OR512VL-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v32i8_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    movl $15, %eax
+; AVX2-NEXT:    vmovd %eax, %xmm1
+; AVX2-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-NEXT:    retq
+;
+; AVX512VLBW-LABEL: shuffle_v32i8_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLBW:       # %bb.0:
+; AVX512VLBW-NEXT:    movl $15, %eax
+; AVX512VLBW-NEXT:    vmovd %eax, %xmm1
+; AVX512VLBW-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VLBW-NEXT:    retq
+;
+; AVX512VLVBMI-SLOW-LABEL: shuffle_v32i8_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-SLOW:       # %bb.0:
+; AVX512VLVBMI-SLOW-NEXT:    movl $15, %eax
+; AVX512VLVBMI-SLOW-NEXT:    vmovd %eax, %xmm1
+; AVX512VLVBMI-SLOW-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
+; AVX512VLVBMI-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VLVBMI-SLOW-NEXT:    retq
+;
+; AVX512VLVBMI-FAST-LABEL: shuffle_v32i8_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VLVBMI-FAST:       # %bb.0:
+; AVX512VLVBMI-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,15,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512VLVBMI-FAST-NEXT:    vpermb %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
diff --git a/test/CodeGen/X86/vector-shuffle-256-v4.ll b/test/CodeGen/X86/vector-shuffle-256-v4.ll
index fc1abdf6b1f..ed281c31d46 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v4.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v4.ll
@@ -937,19 +937,11 @@ define <4 x i64> @shuffle_v4i64_0412(<4 x i64> %a, <4 x i64> %b) {
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX2-NEXT:    retq
 ;
-; AVX512VL-SLOW-LABEL: shuffle_v4i64_0412:
-; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpbroadcastq %xmm1, %xmm1
-; AVX512VL-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,1,2]
-; AVX512VL-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX512VL-SLOW-NEXT:    retq
-;
-; AVX512VL-FAST-LABEL: shuffle_v4i64_0412:
-; AVX512VL-FAST:       # %bb.0:
-; AVX512VL-FAST-NEXT:    vpbroadcastq %xmm1, %xmm1
-; AVX512VL-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,5,1,2]
-; AVX512VL-FAST-NEXT:    vpermt2q %ymm1, %ymm2, %ymm0
-; AVX512VL-FAST-NEXT:    retq
+; AVX512VL-LABEL: shuffle_v4i64_0412:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,4,1,2]
+; AVX512VL-NEXT:    vpermt2q %ymm1, %ymm2, %ymm0
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 4, i32 1, i32 2>
   ret <4 x i64> %shuffle
 }
diff --git a/test/CodeGen/X86/vector-shuffle-256-v8.ll b/test/CodeGen/X86/vector-shuffle-256-v8.ll
index 2abf0ef58fe..74c477300d6 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v8.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v8.ll
@@ -28,11 +28,29 @@ define <8 x float> @shuffle_v8f32_00000010(<8 x float> %a, <8 x float> %b) {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8f32_00000010:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8f32_00000010:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8f32_00000010:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,0,0,1,0]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8f32_00000010:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8f32_00000010:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,0,0,1,0]
+; AVX512VL-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0>
   ret <8 x float> %shuffle
 }
@@ -45,11 +63,29 @@ define <8 x float> @shuffle_v8f32_00000200(<8 x float> %a, <8 x float> %b) {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8f32_00000200:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8f32_00000200:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8f32_00000200:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,0,2,0,0]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8f32_00000200:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8f32_00000200:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,0,2,0,0]
+; AVX512VL-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0>
   ret <8 x float> %shuffle
 }
@@ -62,11 +98,29 @@ define <8 x float> @shuffle_v8f32_00003000(<8 x float> %a, <8 x float> %b) {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8f32_00003000:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8f32_00003000:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8f32_00003000:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,3,0,0,0]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8f32_00003000:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8f32_00003000:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,3,0,0,0]
+; AVX512VL-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 3, i32 0, i32 0, i32 0>
   ret <8 x float> %shuffle
 }
@@ -175,11 +229,29 @@ define <8 x float> @shuffle_v8f32_00001111(<8 x float> %a, <8 x float> %b) {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8f32_00001111:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8f32_00001111:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8f32_00001111:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,1,1,1,1]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8f32_00001111:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8f32_00001111:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,1,1,1,1]
+; AVX512VL-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1>
   ret <8 x float> %shuffle
 }
@@ -292,20 +364,29 @@ define <8 x float> @shuffle_v8f32_08991abb(<8 x float> %a, <8 x float> %b) {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v8f32_08991abb:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vmovaps {{.*#+}} ymm2 = <u,0,1,1,u,2,3,3>
-; AVX2-NEXT:    vpermps %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8f32_08991abb:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} ymm2 = <u,0,1,1,u,2,3,3>
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8f32_08991abb:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = <0,1,1,3,1,3,u,u>
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = <u,0,1,1,u,2,3,3>
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v8f32_08991abb:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,1,1,3]
-; AVX512VL-NEXT:    vmovaps {{.*#+}} ymm0 = [8,0,1,1,10,2,3,3]
-; AVX512VL-NEXT:    vpermi2ps %ymm2, %ymm1, %ymm0
+; AVX512VL-NEXT:    vmovaps {{.*#+}} ymm2 = [8,0,1,1,9,2,3,3]
+; AVX512VL-NEXT:    vpermi2ps %ymm0, %ymm1, %ymm2
+; AVX512VL-NEXT:    vmovaps %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 8, i32 9, i32 9, i32 1, i32 10, i32 11, i32 11>
   ret <8 x float> %shuffle
@@ -344,26 +425,26 @@ define <8 x float> @shuffle_v8f32_09ab1def(<8 x float> %a, <8 x float> %b) {
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v8f32_09ab1def:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8f32_09ab1def:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT:    retq
 ;
-; AVX512VL-SLOW-LABEL: shuffle_v8f32_09ab1def:
-; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; AVX512VL-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX512VL-SLOW-NEXT:    retq
+; AVX2-FAST-LABEL: shuffle_v8f32_09ab1def:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = <0,1,1,3,1,3,u,u>
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT:    retq
 ;
-; AVX512VL-FAST-LABEL: shuffle_v8f32_09ab1def:
-; AVX512VL-FAST:       # %bb.0:
-; AVX512VL-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,1,1,3]
-; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm0 = [8,1,2,3,10,5,6,7]
-; AVX512VL-FAST-NEXT:    vpermi2ps %ymm2, %ymm1, %ymm0
-; AVX512VL-FAST-NEXT:    retq
+; AVX512VL-LABEL: shuffle_v8f32_09ab1def:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vmovaps {{.*#+}} ymm2 = [8,1,2,3,9,5,6,7]
+; AVX512VL-NEXT:    vpermi2ps %ymm0, %ymm1, %ymm2
+; AVX512VL-NEXT:    vmovaps %ymm2, %ymm0
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 1, i32 13, i32 14, i32 15>
   ret <8 x float> %shuffle
 }
@@ -740,11 +821,29 @@ define <8 x float> @shuffle_v8f32_32103210(<8 x float> %a, <8 x float> %b) {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8f32_32103210:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8f32_32103210:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8f32_32103210:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [3,2,1,0,3,2,1,0]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8f32_32103210:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8f32_32103210:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [3,2,1,0,3,2,1,0]
+; AVX512VL-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0>
   ret <8 x float> %shuffle
 }
@@ -1097,11 +1196,29 @@ define <8 x i32> @shuffle_v8i32_00000010(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8i32_00000010:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8i32_00000010:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i32_00000010:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,0,0,1,0]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i32_00000010:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i32_00000010:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,0,0,1,0]
+; AVX512VL-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0>
   ret <8 x i32> %shuffle
 }
@@ -1114,11 +1231,29 @@ define <8 x i32> @shuffle_v8i32_00000200(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8i32_00000200:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8i32_00000200:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i32_00000200:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,0,2,0,0]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i32_00000200:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i32_00000200:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,0,2,0,0]
+; AVX512VL-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0>
   ret <8 x i32> %shuffle
 }
@@ -1131,11 +1266,29 @@ define <8 x i32> @shuffle_v8i32_00003000(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8i32_00003000:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8i32_00003000:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i32_00003000:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,3,0,0,0]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i32_00003000:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i32_00003000:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,3,0,0,0]
+; AVX512VL-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 3, i32 0, i32 0, i32 0>
   ret <8 x i32> %shuffle
 }
@@ -1249,11 +1402,29 @@ define <8 x i32> @shuffle_v8i32_00001111(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8i32_00001111:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8i32_00001111:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i32_00001111:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,1,1,1,1]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i32_00001111:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i32_00001111:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,0,1,1,1,1]
+; AVX512VL-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1>
   ret <8 x i32> %shuffle
 }
@@ -1897,11 +2068,29 @@ define <8 x i32> @shuffle_v8i32_32103210(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8i32_32103210:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8i32_32103210:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i32_32103210:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [3,2,1,0,3,2,1,0]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i32_32103210:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i32_32103210:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [3,2,1,0,3,2,1,0]
+; AVX512VL-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0>
   ret <8 x i32> %shuffle
 }
diff --git a/test/CodeGen/X86/vector-shuffle-avx512.ll b/test/CodeGen/X86/vector-shuffle-avx512.ll
index 07d9107009b..0da2bd82063 100644
--- a/test/CodeGen/X86/vector-shuffle-avx512.ll
+++ b/test/CodeGen/X86/vector-shuffle-avx512.ll
@@ -541,35 +541,33 @@ define <8 x float> @expand14(<4 x float> %a) {
 define <8 x float> @expand15(<4 x float> %a) {
 ; SKX64-LABEL: expand15:
 ; SKX64:       # %bb.0:
-; SKX64-NEXT:    vmovaps {{.*#+}} xmm1 = [0,2,0,0]
-; SKX64-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,1,1,3]
-; SKX64-NEXT:    vmovaps {{.*#+}} ymm0 = [0,1,8,3,10,3,2,3]
-; SKX64-NEXT:    vpermi2ps %ymm2, %ymm1, %ymm0
+; SKX64-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,1,1,3]
+; SKX64-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,8,3,10,5,6,7]
+; SKX64-NEXT:    vbroadcastss {{.*#+}} ymm0 = [2,2,2,2,2,2,2,2]
+; SKX64-NEXT:    vpermt2ps %ymm1, %ymm2, %ymm0
 ; SKX64-NEXT:    retq
 ;
 ; KNL64-LABEL: expand15:
 ; KNL64:       # %bb.0:
-; KNL64-NEXT:    vmovaps {{.*#+}} xmm1 = [0,2,0,0]
-; KNL64-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,1]
 ; KNL64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
 ; KNL64-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,3]
+; KNL64-NEXT:    vbroadcastss {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2]
 ; KNL64-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3],ymm0[4],ymm1[5,6,7]
 ; KNL64-NEXT:    retq
 ;
 ; SKX32-LABEL: expand15:
 ; SKX32:       # %bb.0:
-; SKX32-NEXT:    vmovaps {{.*#+}} xmm1 = [0,2,0,0]
-; SKX32-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,1,1,3]
-; SKX32-NEXT:    vmovaps {{.*#+}} ymm0 = [0,1,8,3,10,3,2,3]
-; SKX32-NEXT:    vpermi2ps %ymm2, %ymm1, %ymm0
+; SKX32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,1,1,3]
+; SKX32-NEXT:    vmovaps {{.*#+}} ymm2 = [0,1,8,3,10,5,6,7]
+; SKX32-NEXT:    vbroadcastss {{.*#+}} ymm0 = [2,2,2,2,2,2,2,2]
+; SKX32-NEXT:    vpermt2ps %ymm1, %ymm2, %ymm0
 ; SKX32-NEXT:    retl
 ;
 ; KNL32-LABEL: expand15:
 ; KNL32:       # %bb.0:
-; KNL32-NEXT:    vmovaps {{.*#+}} xmm1 = [0,2,0,0]
-; KNL32-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,1]
 ; KNL32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
 ; KNL32-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,3]
+; KNL32-NEXT:    vbroadcastss {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2]
 ; KNL32-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3],ymm0[4],ymm1[5,6,7]
 ; KNL32-NEXT:    retl
    %addV = fadd <4 x float> <float 0.0,float 1.0,float 2.0,float 0.0> , <float 0.0,float 1.0,float 2.0,float 0.0>
diff --git a/test/CodeGen/X86/vector-shuffle-v1.ll b/test/CodeGen/X86/vector-shuffle-v1.ll
index 5283d5be395..19c3deb31e7 100644
--- a/test/CodeGen/X86/vector-shuffle-v1.ll
+++ b/test/CodeGen/X86/vector-shuffle-v1.ll
@@ -540,9 +540,8 @@ define i8 @shuf8i1_10_2_9_u_3_u_2_u(i8 %a) {
 ; AVX512VL-NEXT:    kmovw %edi, %k1
 ; AVX512VL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
 ; AVX512VL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3]
 ; AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [8,1,10,3,0,1,2,3]
+; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [8,2,10,3,3,2,2,3]
 ; AVX512VL-NEXT:    vpermi2d %ymm1, %ymm0, %ymm2
 ; AVX512VL-NEXT:    vpslld $31, %ymm2, %ymm0
 ; AVX512VL-NEXT:    vptestmd %ymm0, %ymm0, %k0
@@ -555,9 +554,8 @@ define i8 @shuf8i1_10_2_9_u_3_u_2_u(i8 %a) {
 ; VL_BW_DQ:       # %bb.0:
 ; VL_BW_DQ-NEXT:    kmovd %edi, %k0
 ; VL_BW_DQ-NEXT:    vpmovm2d %k0, %ymm0
-; VL_BW_DQ-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3]
 ; VL_BW_DQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; VL_BW_DQ-NEXT:    vmovdqa {{.*#+}} ymm2 = [8,1,10,3,0,1,2,3]
+; VL_BW_DQ-NEXT:    vmovdqa {{.*#+}} ymm2 = [8,2,10,3,3,2,2,3]
 ; VL_BW_DQ-NEXT:    vpermi2d %ymm1, %ymm0, %ymm2
 ; VL_BW_DQ-NEXT:    vpmovd2m %ymm2, %k0
 ; VL_BW_DQ-NEXT:    kmovd %k0, %eax
@@ -677,9 +675,7 @@ define i8 @shuf8i1_9_6_1_10_3_7_7_0(i8 %a) {
 ; AVX512VL-NEXT:    kmovw %edi, %k1
 ; AVX512VL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
 ; AVX512VL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,2]
-; AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3],ymm1[4,5,6,7]
+; AVX512VL-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[4,5,6,7],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[8,9,10,11],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX512VL-NEXT:    vptestmd %ymm0, %ymm0, %k0
 ; AVX512VL-NEXT:    kmovw %k0, %eax
 ; AVX512VL-NEXT:    # kill: def $al killed $al killed $eax
@@ -690,9 +686,7 @@ define i8 @shuf8i1_9_6_1_10_3_7_7_0(i8 %a) {
 ; VL_BW_DQ:       # %bb.0:
 ; VL_BW_DQ-NEXT:    kmovd %edi, %k0
 ; VL_BW_DQ-NEXT:    vpmovm2d %k0, %ymm0
-; VL_BW_DQ-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,2]
-; VL_BW_DQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; VL_BW_DQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3],ymm1[4,5,6,7]
+; VL_BW_DQ-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[4,5,6,7],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[8,9,10,11],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; VL_BW_DQ-NEXT:    vpmovd2m %ymm0, %k0
 ; VL_BW_DQ-NEXT:    kmovd %k0, %eax
 ; VL_BW_DQ-NEXT:    # kill: def $al killed $al killed $eax