[llvm] 3c42372 - [X86] combineLoad - improve constant pool matches by ignoring undef elements
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Thu Dec 14 07:20:20 PST 2023
Author: Simon Pilgrim
Date: 2023-12-14T15:13:39Z
New Revision: 3c423722cfd75bf293e42001145fe9ebad53c522
URL: https://github.com/llvm/llvm-project/commit/3c423722cfd75bf293e42001145fe9ebad53c522
DIFF: https://github.com/llvm/llvm-project/commit/3c423722cfd75bf293e42001145fe9ebad53c522.diff
LOG: [X86] combineLoad - improve constant pool matches by ignoring undef elements
When trying to share constant pool entries, we can ignore the undef elements of the entry that is being removed
Added:
Modified:
llvm/lib/Target/X86/X86ISelLowering.cpp
llvm/test/CodeGen/X86/splat-for-size.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-8.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index e779a3d3dab6c2..075a49df9da9a5 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -49951,6 +49951,18 @@ static SDValue combineLoad(SDNode *N, SelectionDAG &DAG,
Extract = DAG.getBitcast(RegVT, Extract);
return DCI.CombineTo(N, Extract, SDValue(User, 1));
}
+ auto MatchingBits = [](const APInt &Undefs, const APInt &UserUndefs,
+ ArrayRef<APInt> Bits, ArrayRef<APInt> UserBits) {
+ if (!UserUndefs.isSubsetOf(Undefs))
+ return false;
+ for (unsigned I = 0, E = Undefs.getBitWidth(); I != E; ++I) {
+ if (Undefs[I])
+ continue;
+ if (Bits[I] != UserBits[I])
+ return false;
+ }
+ return true;
+ };
if (User->getOpcode() == X86ISD::VBROADCAST_LOAD &&
getTargetConstantFromBasePtr(Ptr)) {
// See if we are loading a constant that has also been broadcast.
@@ -49961,7 +49973,7 @@ static SDValue combineLoad(SDNode *N, SelectionDAG &DAG,
UserBits)) {
UserUndefs = UserUndefs.trunc(Undefs.getBitWidth());
UserBits.truncate(Bits.size());
- if (Bits == UserBits && UserUndefs.isSubsetOf(Undefs)) {
+ if (MatchingBits(Undefs, UserUndefs, Bits, UserBits)) {
SDValue Extract = extractSubVector(
SDValue(User, 0), 0, DAG, SDLoc(N), RegVT.getSizeInBits());
Extract = DAG.getBitcast(RegVT, Extract);
@@ -49985,7 +49997,7 @@ static SDValue combineLoad(SDNode *N, SelectionDAG &DAG,
UserBits)) {
UserUndefs = UserUndefs.trunc(Undefs.getBitWidth());
UserBits.truncate(Bits.size());
- if (Bits == UserBits && UserUndefs.isSubsetOf(Undefs)) {
+ if (MatchingBits(Undefs, UserUndefs, Bits, UserBits)) {
SDValue Extract = extractSubVector(
SDValue(User, 0), 0, DAG, SDLoc(N), RegVT.getSizeInBits());
Extract = DAG.getBitcast(RegVT, Extract);
diff --git a/llvm/test/CodeGen/X86/splat-for-size.ll b/llvm/test/CodeGen/X86/splat-for-size.ll
index 6325f083d59bd1..76a95e25045700 100644
--- a/llvm/test/CodeGen/X86/splat-for-size.ll
+++ b/llvm/test/CodeGen/X86/splat-for-size.ll
@@ -387,10 +387,9 @@ define <32 x i8> @splat_v32i8_pgso(<32 x i8> %x) !prof !14 {
define <8 x i64> @pr23259() #1 {
; AVX-LABEL: pr23259:
; AVX: # %bb.0: # %entry
-; AVX-NEXT: vbroadcastsd {{.*#+}} ymm0 = [1,1,1,1]
-; AVX-NEXT: vblendps {{.*#+}} xmm1 = mem[0,1],xmm0[2,3]
-; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT: vbroadcastsd {{.*#+}} ymm1 = [1,1,1,1]
+; AVX-NEXT: vblendps {{.*#+}} xmm0 = mem[0,1],xmm1[2,3]
+; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX-NEXT: retq
;
; AVX2-LABEL: pr23259:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
index 8aed554ff58f54..384e9cf36cf85d 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
@@ -603,9 +603,10 @@ define void @load_i16_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpblendvb %ymm6, %ymm1, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-ONLY-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3],ymm0[4,5],ymm1[6],ymm0[7],ymm1[8],ymm0[9,10],ymm1[11],ymm0[12,13],ymm1[14],ymm0[15]
-; AVX2-ONLY-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX2-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX2-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,0,1,6,7,12,13,2,3,8,9,14,15]
+; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} ymm1 = <4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX2-ONLY-NEXT: vpshufb %ymm1, %ymm0, %ymm0
+; AVX2-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX2-ONLY-NEXT: vpshufb %xmm1, %xmm3, %xmm1
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-ONLY-NEXT: vmovdqa %ymm2, (%rsi)
@@ -646,9 +647,10 @@ define void @load_i16_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-NEXT: vpternlogq $202, %ymm2, %ymm1, %ymm0
; AVX512F-NEXT: vmovdqa 16(%rdi), %xmm1
; AVX512F-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3],ymm0[4,5],ymm1[6],ymm0[7],ymm1[8],ymm0[9,10],ymm1[11],ymm0[12,13],ymm1[14],ymm0[15]
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpblendw {{.*#+}} xmm1 = xmm5[0],xmm4[1],xmm5[2,3],xmm4[4],xmm5[5,6],xmm4[7]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,0,1,6,7,12,13,2,3,8,9,14,15]
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm1 = <4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-NEXT: vpshufb %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT: vpblendw {{.*#+}} xmm2 = xmm5[0],xmm4[1],xmm5[2,3],xmm4[4],xmm5[5,6],xmm4[7]
+; AVX512F-NEXT: vpshufb %xmm1, %xmm2, %xmm1
; AVX512F-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512F-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX512F-NEXT: vmovdqa %ymm3, (%rsi)
@@ -1111,22 +1113,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX512F-LABEL: load_i16_stride3_vf32:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
; AVX512F-NEXT: vmovdqa 128(%rdi), %ymm5
; AVX512F-NEXT: vmovdqa 160(%rdi), %ymm6
-; AVX512F-NEXT: vmovdqa %ymm0, %ymm1
-; AVX512F-NEXT: vpternlogq $202, %ymm5, %ymm6, %ymm1
-; AVX512F-NEXT: vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512F-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512F-NEXT: vmovdqa 112(%rdi), %xmm1
+; AVX512F-NEXT: vmovdqa %ymm1, %ymm0
+; AVX512F-NEXT: vpternlogq $202, %ymm5, %ymm6, %ymm0
+; AVX512F-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512F-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
+; AVX512F-NEXT: vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512F-NEXT: vmovdqa 112(%rdi), %xmm0
; AVX512F-NEXT: vmovdqa 96(%rdi), %xmm2
-; AVX512F-NEXT: vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512F-NEXT: vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
; AVX512F-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
; AVX512F-NEXT: vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX512F-NEXT: vmovdqa (%rdi), %ymm8
; AVX512F-NEXT: vmovdqa 32(%rdi), %ymm9
-; AVX512F-NEXT: vmovdqa %ymm0, %ymm3
+; AVX512F-NEXT: vmovdqa %ymm1, %ymm3
; AVX512F-NEXT: vpternlogq $202, %ymm9, %ymm8, %ymm3
; AVX512F-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
; AVX512F-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -1140,48 +1142,49 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-NEXT: vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
; AVX512F-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
; AVX512F-NEXT: vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512F-NEXT: vmovdqa %ymm0, %ymm10
+; AVX512F-NEXT: vmovdqa %ymm1, %ymm10
; AVX512F-NEXT: vpternlogq $202, %ymm6, %ymm5, %ymm10
; AVX512F-NEXT: vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
; AVX512F-NEXT: vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm10 = ymm10[2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512F-NEXT: vpblendw {{.*#+}} xmm11 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[2,3,8,9,14,15,4,5,10,11,u,u,u,u,u,u]
-; AVX512F-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2,3,4],xmm10[5,6,7]
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm11 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512F-NEXT: vmovdqa %ymm11, %ymm12
-; AVX512F-NEXT: vpternlogq $202, %ymm8, %ymm9, %ymm12
-; AVX512F-NEXT: vpermq {{.*#+}} ymm13 = ymm12[2,3,0,1]
-; AVX512F-NEXT: vpblendw {{.*#+}} ymm12 = ymm12[0,1],ymm13[2],ymm12[3,4],ymm13[5],ymm12[6,7,8,9],ymm13[10],ymm12[11,12],ymm13[13],ymm12[14,15]
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm12 = ymm12[2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512F-NEXT: vpblendw {{.*#+}} xmm13 = xmm3[0,1],xmm4[2],xmm3[3,4],xmm4[5],xmm3[6,7]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[u,u,u,u,u,u,4,5,10,11,0,1,6,7,12,13]
-; AVX512F-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
-; AVX512F-NEXT: vpblendw {{.*#+}} ymm13 = ymm12[0,1,2],ymm13[3,4,5,6,7],ymm12[8,9,10],ymm13[11,12,13,14,15]
-; AVX512F-NEXT: vpshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,5,6,7,4]
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
-; AVX512F-NEXT: vinserti64x4 $1, %ymm10, %zmm12, %zmm10
-; AVX512F-NEXT: vpternlogq $202, %ymm5, %ymm6, %ymm11
-; AVX512F-NEXT: vpermq {{.*#+}} ymm5 = ymm11[2,3,0,1]
-; AVX512F-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0],ymm11[1,2],ymm5[3],ymm11[4,5],ymm5[6],ymm11[7],ymm5[8],ymm11[9,10],ymm5[11],ymm11[12,13],ymm5[14],ymm11[15]
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512F-NEXT: vpshufb %ymm11, %ymm10, %ymm10
+; AVX512F-NEXT: vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
+; AVX512F-NEXT: vpshufb %xmm11, %xmm12, %xmm12
+; AVX512F-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512F-NEXT: vmovdqa %ymm12, %ymm13
+; AVX512F-NEXT: vpternlogq $202, %ymm8, %ymm9, %ymm13
+; AVX512F-NEXT: vpermq {{.*#+}} ymm14 = ymm13[2,3,0,1]
+; AVX512F-NEXT: vpblendw {{.*#+}} ymm13 = ymm13[0,1],ymm14[2],ymm13[3,4],ymm14[5],ymm13[6,7,8,9],ymm14[10],ymm13[11,12],ymm14[13],ymm13[14,15]
+; AVX512F-NEXT: vpshufb {{.*#+}} ymm13 = ymm13[2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
+; AVX512F-NEXT: vpblendw {{.*#+}} xmm14 = xmm3[0,1],xmm4[2],xmm3[3,4],xmm4[5],xmm3[6,7]
+; AVX512F-NEXT: vpshufb %xmm11, %xmm14, %xmm11
+; AVX512F-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
+; AVX512F-NEXT: vpblendw {{.*#+}} ymm11 = ymm13[0,1,2],ymm11[3,4,5,6,7],ymm13[8,9,10],ymm11[11,12,13,14,15]
+; AVX512F-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,5,6,7,4]
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm11 = ymm13[0,1,2,3],ymm11[4,5,6,7]
+; AVX512F-NEXT: vinserti64x4 $1, %ymm10, %zmm11, %zmm10
+; AVX512F-NEXT: vpternlogq $202, %ymm5, %ymm6, %ymm12
+; AVX512F-NEXT: vpermq {{.*#+}} ymm5 = ymm12[2,3,0,1]
+; AVX512F-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
; AVX512F-NEXT: vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
; AVX512F-NEXT: vpshufb %ymm6, %ymm5, %ymm5
-; AVX512F-NEXT: vpternlogq $202, %ymm8, %ymm9, %ymm0
-; AVX512F-NEXT: vpermq {{.*#+}} ymm8 = ymm0[2,3,0,1]
-; AVX512F-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm0[1,2],ymm8[3],ymm0[4,5],ymm8[6],ymm0[7],ymm8[8],ymm0[9,10],ymm8[11],ymm0[12,13],ymm8[14],ymm0[15]
-; AVX512F-NEXT: vpshufb %ymm6, %ymm0, %ymm0
+; AVX512F-NEXT: vpternlogq $202, %ymm8, %ymm9, %ymm1
+; AVX512F-NEXT: vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
+; AVX512F-NEXT: vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
+; AVX512F-NEXT: vpshufb %ymm6, %ymm1, %ymm1
; AVX512F-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,0,1,6,7,12,13,2,3,8,9,14,15]
+; AVX512F-NEXT: vpshufb %xmm6, %xmm3, %xmm3
; AVX512F-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
-; AVX512F-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm1
-; AVX512F-NEXT: vextracti32x4 $2, %zmm1, %xmm1
-; AVX512F-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
-; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
+; AVX512F-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512F-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT: vextracti32x4 $2, %zmm0, %xmm0
+; AVX512F-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX512F-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
; AVX512F-NEXT: vmovdqa64 %zmm7, (%rsi)
; AVX512F-NEXT: vmovdqa64 %zmm10, (%rdx)
; AVX512F-NEXT: vmovdqa64 %zmm0, (%rcx)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
index f5399e791f6e84..30f96e0b3e1492 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
@@ -3184,176 +3184,192 @@ define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vmovdqa 256(%rdi), %ymm0
; AVX512F-SLOW-NEXT: vmovdqa 288(%rdi), %ymm1
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0,1],ymm0[2],ymm1[3],ymm0[4],ymm1[5,6],ymm0[7],ymm1[8,9],ymm0[10],ymm1[11],ymm0[12],ymm1[13,14],ymm0[15]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm17
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3,4],xmm2[5,6,7]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,4,5,14,15,8,9,2,3,12,13,6,7]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512F-SLOW-NEXT: vmovdqa 192(%rdi), %ymm7
-; AVX512F-SLOW-NEXT: vmovdqa 224(%rdi), %ymm8
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm8[0],ymm7[1],ymm8[2,3],ymm7[4],ymm8[5],ymm7[6],ymm8[7,8],ymm7[9],ymm8[10,11],ymm7[12],ymm8[13],ymm7[14],ymm8[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0],ymm3[1,2],ymm4[3],ymm3[4,5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[u,u,u,u,u,u,u,u,8,9,2,3,12,13,6,7,16,17,26,27,26,27,30,31,24,25,18,19,28,29,22,23]
-; AVX512F-SLOW-NEXT: vmovdqa 176(%rdi), %xmm9
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm9[3,1,2,3]
-; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa 160(%rdi), %xmm10
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm10[0,2,2,3]
-; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,3,2,3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm3[0,1,2,3,4],ymm2[5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa (%rdi), %ymm5
-; AVX512F-SLOW-NEXT: vmovdqa 32(%rdi), %ymm6
-; AVX512F-SLOW-NEXT: vmovdqa 64(%rdi), %ymm2
-; AVX512F-SLOW-NEXT: vmovdqa 96(%rdi), %ymm3
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm11 = ymm2[0],ymm3[1,2],ymm2[3],ymm3[4],ymm2[5],ymm3[6,7],ymm2[8],ymm3[9,10],ymm2[11],ymm3[12],ymm2[13],ymm3[14,15]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm4
+; AVX512F-SLOW-NEXT: vmovdqa 192(%rdi), %ymm3
+; AVX512F-SLOW-NEXT: vmovdqa 224(%rdi), %ymm9
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm9[0],ymm3[1],ymm9[2,3],ymm3[4],ymm9[5],ymm3[6],ymm9[7,8],ymm3[9],ymm9[10,11],ymm3[12],ymm9[13],ymm3[14],ymm9[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0],ymm5[1,2],ymm6[3],ymm5[4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,8,9,2,3,12,13,6,7,16,17,26,27,26,27,30,31,24,25,18,19,28,29,22,23>
+; AVX512F-SLOW-NEXT: vpshufb %ymm7, %ymm5, %ymm5
+; AVX512F-SLOW-NEXT: vmovdqa 176(%rdi), %xmm1
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm8 = xmm1[3,1,2,3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm20
+; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[2,1,2,3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa 160(%rdi), %xmm6
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm10 = xmm6[0,2,2,3]
+; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,3,2,3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm10[0],xmm8[0],xmm10[1],xmm8[1]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm5[0,1,2,3,4],ymm4[5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa (%rdi), %ymm8
+; AVX512F-SLOW-NEXT: vmovdqa 32(%rdi), %ymm10
+; AVX512F-SLOW-NEXT: vmovdqa 64(%rdi), %ymm4
+; AVX512F-SLOW-NEXT: vmovdqa 96(%rdi), %ymm5
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm11 = ymm4[0],ymm5[1,2],ymm4[3],ymm5[4],ymm4[5],ymm5[6,7],ymm4[8],ymm5[9,10],ymm4[11],ymm5[12],ymm4[13],ymm5[14,15]
; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm12 = ymm11[2,3,0,1]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm12[5],ymm11[6],ymm12[7]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm11 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm11[6,7,16,17,26,27,20,21,30,31,24,25],zero,zero,zero,zero,zero,zero
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm12 = ymm5[0],ymm6[1],ymm5[2,3],ymm6[4],ymm5[5],ymm6[6],ymm5[7,8],ymm6[9],ymm5[10,11],ymm6[12],ymm5[13],ymm6[14],ymm5[15]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm12 = ymm8[0],ymm10[1],ymm8[2,3],ymm10[4],ymm8[5],ymm10[6],ymm8[7,8],ymm10[9],ymm8[10,11],ymm10[12],ymm8[13],ymm10[14],ymm8[15]
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm12, %xmm13
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0],xmm13[1,2,3],xmm12[4,5],xmm13[6,7]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm12 = ymm12[0,1,10,11,4,5,14,15,8,9,2,3,12,13],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm12[u,u,u,u,u,u]
-; AVX512F-SLOW-NEXT: vpor %ymm11, %ymm12, %ymm13
-; AVX512F-SLOW-NEXT: vmovdqa 144(%rdi), %xmm12
-; AVX512F-SLOW-NEXT: vmovdqa 128(%rdi), %xmm11
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm14 = xmm11[0],xmm12[1],xmm11[2,3]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm14 = xmm14[u,u,u,u,u,u,u,u,u,u,2,3,12,13,6,7]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm15
-; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm17 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
-; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm13, %zmm17, %zmm15
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm15, %zmm16
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm13 = ymm7[0],ymm8[1],ymm7[2,3],ymm8[4],ymm7[5],ymm8[6],ymm7[7,8],ymm8[9],ymm7[10,11],ymm8[12],ymm7[13],ymm8[14],ymm7[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm13, %xmm15
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0],ymm15[1],ymm13[2],ymm15[3],ymm13[4,5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm13 = ymm13[2,3,12,13,0,1,0,1,10,11,4,5,14,15,8,9,18,19,28,29,16,17,16,17,26,27,20,21,30,31,24,25]
-; AVX512F-SLOW-NEXT: vpsrlq $48, %xmm9, %xmm15
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm10[0,3,2,3]
-; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2],xmm13[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm13[4,5,6,7]
+; AVX512F-SLOW-NEXT: vpor %ymm11, %ymm12, %ymm15
+; AVX512F-SLOW-NEXT: vmovdqa 144(%rdi), %xmm11
+; AVX512F-SLOW-NEXT: vmovdqa 128(%rdi), %xmm12
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm13 = xmm12[0],xmm11[1],xmm12[2,3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm7, %xmm13, %xmm7
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
+; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm18 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
+; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm15, %zmm18, %zmm7
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm14, %zmm7, %zmm16
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm3[0],ymm9[1],ymm3[2,3],ymm9[4],ymm3[5],ymm9[6],ymm3[7,8],ymm9[9],ymm3[10,11],ymm9[12],ymm3[13],ymm9[14],ymm3[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm14
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0],ymm14[1],ymm7[2],ymm14[3],ymm7[4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm14 = [2,3,12,13,0,1,0,1,10,11,4,5,14,15,8,9,18,19,28,29,16,17,16,17,26,27,20,21,30,31,24,25]
+; AVX512F-SLOW-NEXT: vpshufb %ymm14, %ymm7, %ymm7
+; AVX512F-SLOW-NEXT: vpsrlq $48, %xmm1, %xmm15
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm13 = xmm6[0,3,2,3]
+; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[1,2,2,3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2],xmm7[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm13[0,1,2,3],ymm7[4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa %ymm0, %ymm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm17, %ymm1
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm13 = ymm0[0],ymm1[1,2],ymm0[3],ymm1[4],ymm0[5],ymm1[6,7],ymm0[8],ymm1[9,10],ymm0[11],ymm1[12],ymm0[13],ymm1[14,15]
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm13, %xmm15
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm13 = xmm15[0],xmm13[1],xmm15[2],xmm13[3]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[u,u,u,u,6,7,0,1,10,11,4,5,14,15,8,9]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm13[5,6,7]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm13 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5],ymm2[6],ymm3[7,8],ymm2[9],ymm3[10,11],ymm2[12],ymm3[13],ymm2[14],ymm3[15]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm13[5,6,7]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm13 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5],ymm4[6],ymm5[7,8],ymm4[9],ymm5[10,11],ymm4[12],ymm5[13],ymm4[14],ymm5[15]
; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm15 = ymm13[2,3,0,1]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm15[5],ymm13[6,7]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm15 = ymm10[0],ymm8[1],ymm10[2],ymm8[3],ymm10[4,5],ymm8[6],ymm10[7,8],ymm8[9],ymm10[10],ymm8[11],ymm10[12,13],ymm8[14],ymm10[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm15, %xmm0
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm15[0,1],xmm0[2,3],xmm15[4,5,6],xmm0[7]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm13 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm13[8,9,18,19,28,29,22,23,16,17,26,27],zero,zero,zero,zero,zero,zero
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm15 = ymm6[0],ymm5[1],ymm6[2],ymm5[3],ymm6[4,5],ymm5[6],ymm6[7,8],ymm5[9],ymm6[10],ymm5[11],ymm6[12,13],ymm5[14],ymm6[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm15, %xmm14
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3],xmm15[4,5,6],xmm14[7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm14 = ymm14[2,3,12,13,6,7,0,1,10,11,4,5,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm14[u,u,u,u,u,u]
-; AVX512F-SLOW-NEXT: vpor %ymm13, %ymm14, %ymm13
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm14 = xmm11[0,1],xmm12[2],xmm11[3]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm14 = xmm14[u,u,u,u,u,u,u,u,u,u,4,5,14,15,8,9]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm13, %zmm17, %zmm14
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm14, %zmm18
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm8[0],ymm7[1],ymm8[2],ymm7[3],ymm8[4,5],ymm7[6],ymm8[7,8],ymm7[9],ymm8[10],ymm7[11],ymm8[12,13],ymm7[14],ymm8[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm14
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0],ymm14[1],ymm4[2,3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[4,5,14,15,12,13,2,3,12,13,6,7,0,1,10,11,20,21,30,31,28,29,18,19,28,29,22,23,16,17,26,27]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm14 = xmm10[0,1,1,3]
-; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,7,6,7]
-; AVX512F-SLOW-NEXT: vpunpckhdq {{.*#+}} xmm14 = xmm14[2],xmm9[2],xmm14[3],xmm9[3]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm14 = xmm14[0,1,2],xmm4[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1,2,3],ymm4[4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm14 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10,11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm14, %xmm15
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm14 = xmm15[0,1,2],xmm14[3,4],xmm15[5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm14 = xmm14[u,u,u,u,8,9,2,3,12,13,6,7,0,1,10,11]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm14[5,6,7]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm14 = ymm6[0,1],ymm5[2],ymm6[3],ymm5[4],ymm6[5,6],ymm5[7],ymm6[8,9],ymm5[10],ymm6[11],ymm5[12],ymm6[13,14],ymm5[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm14, %xmm15
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm14 = xmm14[0,1,2],xmm15[3,4],xmm14[5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm14 = xmm14[4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm15 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5],ymm3[6],ymm2[7,8],ymm3[9],ymm2[10,11],ymm3[12],ymm2[13],ymm3[14],ymm2[15]
-; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm13 = ymm15[2,3,0,1]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5],ymm13[6],ymm15[7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,0,1,10,11,20,21,30,31,24,25,18,19,28,29,26,27,16,17,26,27]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm14 = xmm12[0],xmm11[1],xmm12[2,3]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm14 = xmm14[u,u,u,u,u,u,u,u,u,u,6,7,0,1,10,11]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm13, %zmm17, %zmm14
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm14, %zmm17
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm8[0,1],ymm7[2],ymm8[3],ymm7[4],ymm8[5,6],ymm7[7],ymm8[8,9],ymm7[10],ymm8[11],ymm7[12],ymm8[13,14],ymm7[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm13
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm13[2],ymm4[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[6,7,6,7,8,9,4,5,14,15,8,9,2,3,12,13,22,23,22,23,24,25,20,21,30,31,24,25,18,19,28,29]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm13 = xmm9[0],xmm10[1],xmm9[2,3]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[6,7,0,1,10,11,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2],xmm4[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm13[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[2,3,12,13,6,7,0,1,10,11,4,5,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[u,u,u,u,u,u]
+; AVX512F-SLOW-NEXT: vpor %ymm0, %ymm13, %ymm0
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm13 = xmm12[0,1],xmm11[2],xmm12[3]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[u,u,u,u,u,u,u,u,u,u,8,9,2,3,12,13]
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm13, %xmm13
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm13, %zmm4
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm13 = ymm5[0],ymm6[1,2],ymm5[3],ymm6[4],ymm5[5],ymm6[6,7],ymm5[8],ymm6[9,10],ymm5[11],ymm6[12],ymm5[13],ymm6[14,15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm13, %xmm15
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm13 = xmm15[0],xmm13[1],xmm15[2],xmm13[3]
+; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm0, %zmm18, %zmm13
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm7, %zmm13, %zmm19
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm9[0],ymm3[1],ymm9[2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7,8],ymm3[9],ymm9[10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm7
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = [4,5,14,15,12,13,2,3,12,13,6,7,0,1,10,11,20,21,30,31,28,29,18,19,28,29,22,23,16,17,26,27]
+; AVX512F-SLOW-NEXT: vpshufb %ymm7, %ymm0, %ymm0
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm7, %ymm21
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm7 = xmm6[0,1,1,3]
+; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm15
+; AVX512F-SLOW-NEXT: vpunpckhdq {{.*#+}} xmm7 = xmm7[2],xmm20[2],xmm7[3],xmm20[3]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2],xmm0[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5],ymm2[6],ymm1[7,8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13],ymm2[14],ymm1[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm13
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm7 = xmm13[0,1,2],xmm7[3,4],xmm13[5,6,7]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,8,9,2,3,12,13,6,7,0,1,10,11]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm0[0,1,2,3,4],ymm7[5,6,7]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm9[0,1],ymm3[2],ymm9[3],ymm3[4],ymm9[5,6],ymm3[7],ymm9[8,9],ymm3[10],ymm9[11],ymm3[12],ymm9[13,14],ymm3[15]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm9, %ymm20
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm13
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm13[2],ymm0[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm13 = [6,7,6,7,8,9,4,5,14,15,8,9,2,3,12,13,22,23,22,23,24,25,20,21,30,31,24,25,18,19,28,29]
+; AVX512F-SLOW-NEXT: vpshufb %ymm13, %ymm0, %ymm0
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm14 = xmm15[0],xmm6[1],xmm15[2,3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm15, %xmm22
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm14 = xmm14[6,7,0,1,10,11,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm14 = xmm14[0,1,2],xmm0[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm14 = xmm11[0,1],xmm12[2],xmm11[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm13, %xmm14, %xmm13
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm13, %zmm0
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm13 = ymm8[0],ymm10[1,2],ymm8[3],ymm10[4],ymm8[5],ymm10[6,7],ymm8[8],ymm10[9,10],ymm8[11],ymm10[12],ymm8[13],ymm10[14,15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm13, %xmm14
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm13 = xmm14[0],xmm13[1],xmm14[2],xmm13[3]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm14 = ymm5[0],ymm4[1],ymm5[2],ymm4[3],ymm5[4,5],ymm4[6],ymm5[7,8],ymm4[9],ymm5[10],ymm4[11],ymm5[12,13],ymm4[14],ymm5[15]
+; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm15 = ymm14[2,3,0,1]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4],ymm14[5],ymm15[6],ymm14[7]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm15 = ymm3[0],ymm2[1],ymm3[2],ymm2[3],ymm3[4,5],ymm2[6],ymm3[7,8],ymm2[9],ymm3[10],ymm2[11],ymm3[12,13],ymm2[14],ymm3[15]
-; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm14 = ymm15[2,3,0,1]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3],ymm14[4],ymm15[5],ymm14[6],ymm15[7]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm14 = ymm14[u,u,u,u,u,u,u,u,u,u,u,u,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm13
-; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm13, %ymm4
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm14 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5],ymm1[6],ymm0[7,8],ymm1[9],ymm0[10,11],ymm1[12],ymm0[13],ymm1[14],ymm0[15]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm14 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5],ymm1[6],ymm2[7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13],ymm1[14],ymm2[15]
+; AVX512F-SLOW-NEXT: vmovdqa %ymm2, %ymm9
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm14, %xmm15
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm14 = xmm14[0],xmm15[1,2,3],xmm14[4,5],xmm15[6,7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm14 = xmm14[u,u,0,1,10,11,4,5,14,15,8,9,2,3,12,13]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm14 = ymm4[0],ymm14[1,2,3,4,5,6,7],ymm4[8],ymm14[9,10,11,12,13,14,15]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm14[4,5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm13, %zmm4
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0],ymm8[1,2],ymm7[3],ymm8[4],ymm7[5],ymm8[6,7],ymm7[8],ymm8[9,10],ymm7[11],ymm8[12],ymm7[13],ymm8[14,15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm8
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm8[0],ymm7[1],ymm8[2],ymm7[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm7 = ymm7[8,9,14,15,4,5,6,7,0,1,10,11,4,5,14,15,24,25,30,31,20,21,22,23,16,17,26,27,20,21,30,31]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm8 = xmm9[0,1],xmm10[2],xmm9[3]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[8,9,2,3,12,13,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1,2],xmm7[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm8 = xmm12[3,1,2,3]
-; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,1,2,1,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm9 = xmm11[0,2,2,3]
-; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[0,1,0,3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm7, %zmm8, %zmm7
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm6[0],ymm5[1],ymm6[2,3],ymm5[4],ymm6[5],ymm5[6],ymm6[7,8],ymm5[9],ymm6[10,11],ymm5[12],ymm6[13],ymm5[14],ymm6[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3,4],xmm6[5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[8,9,2,3,12,13,6,7,0,1,10,11,u,u,u,u]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3],ymm2[4],ymm3[5,6],ymm2[7],ymm3[8,9],ymm2[10],ymm3[11],ymm2[12],ymm3[13,14],ymm2[15]
-; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6],ymm3[7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,20,21,22,23,20,21,30,31]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm13
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm14[u,u,0,1,10,11,4,5,14,15,8,9,2,3,12,13]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm13, %ymm14
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm14[0],ymm0[1,2,3,4,5,6,7],ymm14[8],ymm0[9,10,11,12,13,14,15]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm13, %zmm17
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm11[3,1,2,3]
+; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm13 = xmm12[0,2,2,3]
+; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[0,1,0,3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm13 = ymm10[0,1],ymm8[2],ymm10[3],ymm8[4],ymm10[5,6],ymm8[7],ymm10[8,9],ymm8[10],ymm10[11],ymm8[12],ymm10[13,14],ymm8[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm13, %xmm14
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2],xmm14[3,4],xmm13[5,6,7]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm14 = ymm4[0],ymm5[1],ymm4[2,3],ymm5[4],ymm4[5],ymm5[6],ymm4[7,8],ymm5[9],ymm4[10,11],ymm5[12],ymm4[13],ymm5[14],ymm4[15]
+; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm15 = ymm14[2,3,0,1]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6],ymm14[7]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm14 = ymm14[u,u,u,u,u,u,u,u,u,u,u,u,0,1,10,11,20,21,30,31,24,25,18,19,28,29,26,27,16,17,26,27]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm11 = xmm11[0],xmm12[1],xmm11[2,3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm21, %ymm2
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm11, %xmm11
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
+; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm13, %zmm18, %zmm11
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm7, %zmm11, %zmm7
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm20, %ymm2
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1,2],ymm3[3],ymm2[4],ymm3[5],ymm2[6,7],ymm3[8],ymm2[9,10],ymm3[11],ymm2[12],ymm3[13],ymm2[14,15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2],ymm2[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm3
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1],xmm6[2],xmm3[3]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[8,9,14,15,4,5,6,7,0,1,10,11,4,5,14,15,24,25,30,31,20,21,22,23,16,17,26,27,20,21,30,31]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[8,9,2,3,12,13,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm2[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm10[0],ymm8[1],ymm10[2,3],ymm8[4],ymm10[5],ymm8[6],ymm10[7,8],ymm8[9],ymm10[10,11],ymm8[12],ymm10[13],ymm8[14],ymm10[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3,4],xmm3[5,6,7]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm5[0,1],ymm4[2],ymm5[3],ymm4[4],ymm5[5,6],ymm4[7],ymm5[8,9],ymm4[10],ymm5[11],ymm4[12],ymm5[13,14],ymm4[15]
+; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4],ymm3[5,6],ymm4[7]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[8,9,2,3,12,13,6,7,0,1,10,11,u,u,u,u]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,20,21,22,23,20,21,30,31]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX512F-SLOW-NEXT: movb $7, %al
; AVX512F-SLOW-NEXT: kmovw %eax, %k1
-; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm2, %zmm0, %zmm7 {%k1}
-; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm7, %ymm2
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4,5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10],ymm0[11],ymm1[12,13],ymm0[14],ymm1[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6],xmm1[7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,2,3,12,13,6,7,0,1,10,11,4,5,14,15]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm7, %zmm0
+; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm2, %zmm0, %zmm0 {%k1}
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm9[1],ymm1[2],ymm9[3],ymm1[4,5],ymm9[6],ymm1[7,8],ymm9[9],ymm1[10],ymm9[11],ymm1[12,13],ymm9[14],ymm1[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6],xmm2[7]
+; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm0, %ymm2
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,2,3,12,13,6,7,0,1,10,11,4,5,14,15]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3,4,5,6,7],ymm2[8],ymm1[9,10,11,12,13,14,15]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512F-SLOW-NEXT: vmovdqa64 %zmm16, (%rsi)
-; AVX512F-SLOW-NEXT: vmovdqa64 %zmm18, (%rdx)
-; AVX512F-SLOW-NEXT: vmovdqa64 %zmm17, (%rcx)
-; AVX512F-SLOW-NEXT: vmovdqa64 %zmm4, (%r8)
+; AVX512F-SLOW-NEXT: vmovdqa64 %zmm19, (%rdx)
+; AVX512F-SLOW-NEXT: vmovdqa64 %zmm7, (%rcx)
+; AVX512F-SLOW-NEXT: vmovdqa64 %zmm17, (%r8)
; AVX512F-SLOW-NEXT: vmovdqa64 %zmm0, (%r9)
; AVX512F-SLOW-NEXT: vzeroupper
; AVX512F-SLOW-NEXT: retq
@@ -3396,104 +3412,103 @@ define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpor %ymm7, %ymm12, %ymm12
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = <0,3,1,u,0,3,5,u>
; AVX512F-FAST-NEXT: vmovdqa 128(%rdi), %ymm7
-; AVX512F-FAST-NEXT: vpermd %ymm7, %ymm13, %ymm14
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = [4,5,2,3,4,5,6,7,8,9,2,3,4,5,10,11,20,21,18,19,20,21,22,23,24,25,18,19,20,21,26,27]
-; AVX512F-FAST-NEXT: vpshufb %ymm13, %ymm14, %ymm14
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm15 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
-; AVX512F-FAST-NEXT: vpternlogq $184, %zmm12, %zmm15, %zmm14
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm14, %zmm16
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm9[0],ymm8[1],ymm9[2,3],ymm8[4],ymm9[5],ymm8[6],ymm9[7,8],ymm8[9],ymm9[10,11],ymm8[12],ymm9[13],ymm8[14],ymm9[15]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <2,u,u,u,4,7,1,6>
-; AVX512F-FAST-NEXT: vpermd %ymm12, %ymm14, %ymm12
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm12[0,1,18,19,20,21,26,27,16,17,30,31],zero,zero,zero,zero,zero,zero
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm14 = ymm11[0],ymm10[1],ymm11[2],ymm10[3],ymm11[4,5],ymm10[6],ymm11[7,8],ymm10[9],ymm11[10],ymm10[11],ymm11[12,13],ymm10[14],ymm11[15]
-; AVX512F-FAST-NEXT: vextracti128 $1, %ymm14, %xmm6
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm14[0,1],xmm6[2,3],xmm14[4,5,6],xmm6[7]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[2,3,12,13,6,7,0,1,10,11,4,5,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[u,u,u,u,u,u]
-; AVX512F-FAST-NEXT: vpor %ymm6, %ymm12, %ymm6
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm12 = <1,3,2,u,1,3,6,u>
-; AVX512F-FAST-NEXT: vpermd %ymm7, %ymm12, %ymm12
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = [0,1,6,7,4,5,6,7,8,9,0,1,6,7,8,9,16,17,22,23,20,21,22,23,24,25,16,17,22,23,24,25]
-; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm12, %ymm12
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm14, %ymm19
-; AVX512F-FAST-NEXT: vpternlogq $184, %zmm6, %zmm15, %zmm12
-; AVX512F-FAST-NEXT: vpsrlq $48, %xmm2, %xmm6
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm14 = xmm3[2,3,12,13,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-FAST-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm14[0],xmm6[0],xmm14[1],xmm6[1]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm14 = ymm4[0],ymm5[1],ymm4[2,3],ymm5[4],ymm4[5],ymm5[6],ymm4[7,8],ymm5[9],ymm4[10,11],ymm5[12],ymm4[13],ymm5[14],ymm4[15]
+; AVX512F-FAST-NEXT: vpermd %ymm7, %ymm13, %ymm13
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = [4,5,2,3,4,5,6,7,8,9,2,3,4,5,10,11,20,21,18,19,20,21,22,23,24,25,18,19,20,21,26,27]
+; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm13, %ymm13
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm18 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
+; AVX512F-FAST-NEXT: vpternlogq $184, %zmm12, %zmm18, %zmm13
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm13, %zmm16
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm11[0],ymm10[1],ymm11[2],ymm10[3],ymm11[4,5],ymm10[6],ymm11[7,8],ymm10[9],ymm11[10],ymm10[11],ymm11[12,13],ymm10[14],ymm11[15]
+; AVX512F-FAST-NEXT: vextracti128 $1, %ymm12, %xmm13
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0,1],xmm13[2,3],xmm12[4,5,6],xmm13[7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <2,3,12,13,6,7,0,1,10,11,4,5,14,15,128,128,128,128,128,128,128,128,128,128,128,128,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm12, %ymm12
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm13 = ymm9[0],ymm8[1],ymm9[2,3],ymm8[4],ymm9[5],ymm8[6],ymm9[7,8],ymm8[9],ymm9[10,11],ymm8[12],ymm9[13],ymm8[14],ymm9[15]
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm17 = <2,u,u,u,4,7,1,6>
+; AVX512F-FAST-NEXT: vpermd %ymm13, %ymm17, %ymm13
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm13 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm13[0,1,18,19,20,21,26,27,16,17,30,31],zero,zero,zero,zero,zero,zero
+; AVX512F-FAST-NEXT: vpor %ymm13, %ymm12, %ymm12
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = <1,3,2,u,1,3,6,u>
+; AVX512F-FAST-NEXT: vpermd %ymm7, %ymm13, %ymm15
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = [0,1,6,7,4,5,6,7,8,9,0,1,6,7,8,9,16,17,22,23,20,21,22,23,24,25,16,17,22,23,24,25]
+; AVX512F-FAST-NEXT: vpshufb %ymm13, %ymm15, %ymm15
+; AVX512F-FAST-NEXT: vpternlogq $184, %zmm12, %zmm18, %zmm15
+; AVX512F-FAST-NEXT: vpshufb %xmm6, %xmm3, %xmm6
+; AVX512F-FAST-NEXT: vpsrlq $48, %xmm2, %xmm12
+; AVX512F-FAST-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm4[0],ymm5[1],ymm4[2,3],ymm5[4],ymm4[5],ymm5[6],ymm4[7,8],ymm5[9],ymm4[10,11],ymm5[12],ymm4[13],ymm5[14],ymm4[15]
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm17 = <0,2,5,7,4,7,u,u>
-; AVX512F-FAST-NEXT: vpermd %ymm14, %ymm17, %ymm14
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm14 = ymm14[2,3,4,5,4,5,0,1,6,7,8,9,14,15,4,5,18,19,20,21,20,21,16,17,22,23,24,25,30,31,20,21]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2],xmm14[3,4,5,6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm14[4,5,6,7]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm14 = ymm0[0],ymm1[1,2],ymm0[3],ymm1[4],ymm0[5],ymm1[6,7],ymm0[8],ymm1[9,10],ymm0[11],ymm1[12],ymm0[13],ymm1[14,15]
+; AVX512F-FAST-NEXT: vpermd %ymm12, %ymm17, %ymm12
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm12 = ymm12[2,3,4,5,4,5,0,1,6,7,8,9,14,15,4,5,18,19,20,21,20,21,16,17,22,23,24,25,30,31,20,21]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2],xmm12[3,4,5,6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm12[4,5,6,7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm0[0],ymm1[1,2],ymm0[3],ymm1[4],ymm0[5],ymm1[6,7],ymm0[8],ymm1[9,10],ymm0[11],ymm1[12],ymm0[13],ymm1[14,15]
; AVX512F-FAST-NEXT: vbroadcasti32x4 {{.*#+}} ymm17 = [1,4,6,3,1,4,6,3]
; AVX512F-FAST-NEXT: # ymm17 = mem[0,1,2,3,0,1,2,3]
-; AVX512F-FAST-NEXT: vpermd %ymm14, %ymm17, %ymm14
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm14 = ymm14[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27,16,17,30,31,24,25]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm14[5,6,7]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm12, %zmm17
+; AVX512F-FAST-NEXT: vpermd %ymm12, %ymm17, %ymm12
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm12 = ymm12[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27,16,17,30,31,24,25]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm12[5,6,7]
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm15, %zmm17
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10,11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
-; AVX512F-FAST-NEXT: vextracti128 $1, %ymm6, %xmm14
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm14[0,1,2],xmm6[3,4],xmm14[5,6,7]
+; AVX512F-FAST-NEXT: vextracti128 $1, %ymm6, %xmm15
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3,4],xmm15[5,6,7]
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u,u,u,8,9,2,3,12,13,6,7,0,1,10,11]
; AVX512F-FAST-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm14 = xmm3[u,u,u,u,u,u,u,u,4,5,14,15,u,u,u,u]
-; AVX512F-FAST-NEXT: vpunpckhdq {{.*#+}} xmm14 = xmm14[2],xmm2[2],xmm14[3],xmm2[3]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm15 = xmm3[u,u,u,u,u,u,u,u,4,5,14,15,u,u,u,u]
+; AVX512F-FAST-NEXT: vpunpckhdq {{.*#+}} xmm15 = xmm15[2],xmm2[2],xmm15[3],xmm2[3]
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm5[0],ymm4[1],ymm5[2],ymm4[3],ymm5[4,5],ymm4[6],ymm5[7,8],ymm4[9],ymm5[10],ymm4[11],ymm5[12,13],ymm4[14],ymm5[15]
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm18 = <0,3,5,2,5,7,u,u>
-; AVX512F-FAST-NEXT: vpermd %ymm12, %ymm18, %ymm12
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm19 = <0,3,5,2,5,7,u,u>
+; AVX512F-FAST-NEXT: vpermd %ymm12, %ymm19, %ymm12
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm12 = ymm12[0,1,6,7,2,3,2,3,4,5,10,11,0,1,14,15,16,17,22,23,18,19,18,19,20,21,26,27,16,17,30,31]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm14 = xmm14[0,1,2],xmm12[3,4,5,6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm14[0,1,2,3],ymm12[4,5,6,7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm15 = xmm15[0,1,2],xmm12[3,4,5,6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm15[0,1,2,3],ymm12[4,5,6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm12[0,1,2,3,4],ymm6[5,6,7]
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm11[0,1],ymm10[2],ymm11[3],ymm10[4],ymm11[5,6],ymm10[7],ymm11[8,9],ymm10[10],ymm11[11],ymm10[12],ymm11[13,14],ymm10[15]
-; AVX512F-FAST-NEXT: vextracti128 $1, %ymm12, %xmm14
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0,1,2],xmm14[3,4],xmm12[5,6,7]
+; AVX512F-FAST-NEXT: vextracti128 $1, %ymm12, %xmm15
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0,1,2],xmm15[3,4],xmm12[5,6,7]
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm12[4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm14 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5],ymm9[6],ymm8[7,8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13],ymm9[14],ymm8[15]
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm18 = <0,2,u,u,5,7,2,4>
-; AVX512F-FAST-NEXT: vpermd %ymm14, %ymm18, %ymm14
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm14 = ymm14[u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,20,21,22,23,16,17,22,23]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2],ymm14[3,4,5,6,7]
-; AVX512F-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm14 = [1,4,6,0,1,4,6,0]
-; AVX512F-FAST-NEXT: # ymm14 = mem[0,1,0,1]
-; AVX512F-FAST-NEXT: vpermd %ymm7, %ymm14, %ymm14
-; AVX512F-FAST-NEXT: vpshufb %ymm13, %ymm14, %ymm13
-; AVX512F-FAST-NEXT: vpternlogq $184, %zmm12, %zmm15, %zmm13
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm13, %zmm13
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm15 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5],ymm9[6],ymm8[7,8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13],ymm9[14],ymm8[15]
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm19 = <0,2,u,u,5,7,2,4>
+; AVX512F-FAST-NEXT: vpermd %ymm15, %ymm19, %ymm15
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm15 = ymm15[u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,20,21,22,23,16,17,22,23]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2],ymm15[3,4,5,6,7]
+; AVX512F-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm15 = [1,4,6,0,1,4,6,0]
+; AVX512F-FAST-NEXT: # ymm15 = mem[0,1,0,1]
+; AVX512F-FAST-NEXT: vpermd %ymm7, %ymm15, %ymm15
+; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm15, %ymm14
+; AVX512F-FAST-NEXT: vpternlogq $184, %zmm12, %zmm18, %zmm14
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm14, %zmm14
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm10[0],ymm11[1,2],ymm10[3],ymm11[4],ymm10[5],ymm11[6,7],ymm10[8],ymm11[9,10],ymm10[11],ymm11[12],ymm10[13],ymm11[14,15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm6, %xmm12
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm12[0],xmm6[1],xmm12[2],xmm6[3]
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u]
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm9[0],ymm8[1],ymm9[2],ymm8[3],ymm9[4,5],ymm8[6],ymm9[7,8],ymm8[9],ymm9[10],ymm8[11],ymm9[12,13],ymm8[14],ymm9[15]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <0,3,u,u,5,0,2,7>
-; AVX512F-FAST-NEXT: vpermd %ymm12, %ymm14, %ymm12
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <0,3,u,u,5,0,2,7>
+; AVX512F-FAST-NEXT: vpermd %ymm12, %ymm15, %ymm12
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm12 = ymm12[u,u,u,u,u,u,u,u,u,u,u,u,2,3,4,5,18,19,20,21,26,27,16,17,30,31,30,31,18,19,20,21]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2],ymm12[3,4,5,6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm2[0],xmm3[1],xmm2[2,3]
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm12[6,7,0,1,10,11,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm14 = ymm5[0,1],ymm4[2],ymm5[3],ymm4[4],ymm5[5,6],ymm4[7],ymm5[8,9],ymm4[10],ymm5[11],ymm4[12],ymm5[13,14],ymm4[15]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <1,3,6,0,5,u,u,u>
-; AVX512F-FAST-NEXT: vpermd %ymm14, %ymm15, %ymm14
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm14 = ymm14[2,3,2,3,4,5,0,1,6,7,8,9,14,15,4,5,18,19,18,19,20,21,16,17,22,23,24,25,30,31,20,21]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0,1,2],xmm14[3,4,5,6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm14[4,5,6,7]
-; AVX512F-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm14 = [2,4,7,0,2,4,7,0]
-; AVX512F-FAST-NEXT: # ymm14 = mem[0,1,0,1]
-; AVX512F-FAST-NEXT: vpermd %ymm7, %ymm14, %ymm14
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, %ymm15
-; AVX512F-FAST-NEXT: vpshufb %ymm15, %ymm14, %ymm14
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm12, %zmm14, %zmm12
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm15 = ymm5[0,1],ymm4[2],ymm5[3],ymm4[4],ymm5[5,6],ymm4[7],ymm5[8,9],ymm4[10],ymm5[11],ymm4[12],ymm5[13,14],ymm4[15]
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm18 = <1,3,6,0,5,u,u,u>
+; AVX512F-FAST-NEXT: vpermd %ymm15, %ymm18, %ymm15
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm15 = ymm15[2,3,2,3,4,5,0,1,6,7,8,9,14,15,4,5,18,19,18,19,20,21,16,17,22,23,24,25,30,31,20,21]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0,1,2],xmm15[3,4,5,6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm15[4,5,6,7]
+; AVX512F-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm15 = [2,4,7,0,2,4,7,0]
+; AVX512F-FAST-NEXT: # ymm15 = mem[0,1,0,1]
+; AVX512F-FAST-NEXT: vpermd %ymm7, %ymm15, %ymm15
+; AVX512F-FAST-NEXT: vpshufb %ymm13, %ymm15, %ymm13
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm12, %zmm13, %zmm12
; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm6, %zmm12
; AVX512F-FAST-NEXT: vextracti64x4 $1, %zmm12, %ymm6
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm14 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5],ymm1[6],ymm0[7,8],ymm1[9],ymm0[10,11],ymm1[12],ymm0[13],ymm1[14],ymm0[15]
-; AVX512F-FAST-NEXT: vextracti128 $1, %ymm14, %xmm15
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm14 = xmm14[0],xmm15[1,2,3],xmm14[4,5],xmm15[6,7]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm14 = xmm14[u,u,0,1,10,11,4,5,14,15,8,9,2,3,12,13]
-; AVX512F-FAST-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm14 = ymm6[0],ymm14[1,2,3,4,5,6,7],ymm6[8],ymm14[9,10,11,12,13,14,15]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm14[4,5,6,7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm13 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5],ymm1[6],ymm0[7,8],ymm1[9],ymm0[10,11],ymm1[12],ymm0[13],ymm1[14],ymm0[15]
+; AVX512F-FAST-NEXT: vextracti128 $1, %ymm13, %xmm15
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0],xmm15[1,2,3],xmm13[4,5],xmm15[6,7]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[u,u,0,1,10,11,4,5,14,15,8,9,2,3,12,13]
+; AVX512F-FAST-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm13 = ymm6[0],ymm13[1,2,3,4,5,6,7],ymm6[8],ymm13[9,10,11,12,13,14,15]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm13[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm12, %zmm6
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm11[0],ymm10[1],ymm11[2,3],ymm10[4],ymm11[5],ymm10[6],ymm11[7,8],ymm10[9],ymm11[10,11],ymm10[12],ymm11[13],ymm10[14],ymm11[15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm10, %xmm11
@@ -3530,7 +3545,7 @@ define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm0
; AVX512F-FAST-NEXT: vmovdqa64 %zmm16, (%rsi)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm17, (%rdx)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm13, (%rcx)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm14, (%rcx)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm6, (%r8)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm0, (%r9)
; AVX512F-FAST-NEXT: vzeroupper
@@ -6686,13 +6701,13 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX512F-SLOW-LABEL: load_i16_stride5_vf64:
; AVX512F-SLOW: # %bb.0:
-; AVX512F-SLOW-NEXT: subq $584, %rsp # imm = 0x248
+; AVX512F-SLOW-NEXT: subq $552, %rsp # imm = 0x228
; AVX512F-SLOW-NEXT: vmovdqa 384(%rdi), %ymm1
; AVX512F-SLOW-NEXT: vmovdqa 416(%rdi), %ymm2
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm2[1,2],ymm1[3],ymm2[4],ymm1[5],ymm2[6,7],ymm1[8],ymm2[9,10],ymm1[11],ymm2[12],ymm1[13],ymm2[14,15]
-; AVX512F-SLOW-NEXT: vmovdqa %ymm2, %ymm6
+; AVX512F-SLOW-NEXT: vmovdqa %ymm2, %ymm11
; AVX512F-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-SLOW-NEXT: vmovdqa %ymm1, %ymm7
+; AVX512F-SLOW-NEXT: vmovdqa %ymm1, %ymm6
; AVX512F-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6],ymm1[7]
@@ -6701,25 +6716,25 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vmovdqa 352(%rdi), %ymm3
; AVX512F-SLOW-NEXT: vmovdqa 320(%rdi), %ymm4
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5],ymm3[6],ymm4[7,8],ymm3[9],ymm4[10,11],ymm3[12],ymm4[13],ymm3[14],ymm4[15]
-; AVX512F-SLOW-NEXT: vmovdqa %ymm4, %ymm8
+; AVX512F-SLOW-NEXT: vmovdqa %ymm4, %ymm7
; AVX512F-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-SLOW-NEXT: vmovdqa %ymm3, %ymm9
+; AVX512F-SLOW-NEXT: vmovdqa %ymm3, %ymm8
; AVX512F-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm3
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0],xmm3[1,2,3],xmm1[4,5],xmm3[6,7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = <0,1,10,11,4,5,14,15,8,9,2,3,12,13,128,128,128,128,128,128,128,128,128,128,128,128,u,u,u,u,u,u>
; AVX512F-SLOW-NEXT: vpshufb %ymm1, %ymm3, %ymm3
; AVX512F-SLOW-NEXT: vporq %ymm2, %ymm3, %ymm19
-; AVX512F-SLOW-NEXT: vmovdqa 192(%rdi), %ymm12
-; AVX512F-SLOW-NEXT: vmovdqa 224(%rdi), %ymm15
-; AVX512F-SLOW-NEXT: vmovdqa 176(%rdi), %xmm13
+; AVX512F-SLOW-NEXT: vmovdqa 192(%rdi), %ymm15
+; AVX512F-SLOW-NEXT: vmovdqa 224(%rdi), %ymm13
+; AVX512F-SLOW-NEXT: vmovdqa 176(%rdi), %xmm12
; AVX512F-SLOW-NEXT: vmovdqa 160(%rdi), %xmm14
; AVX512F-SLOW-NEXT: vmovdqa (%rdi), %ymm4
; AVX512F-SLOW-NEXT: vmovdqa 32(%rdi), %ymm5
; AVX512F-SLOW-NEXT: vmovdqa 64(%rdi), %ymm3
-; AVX512F-SLOW-NEXT: vmovdqa 96(%rdi), %ymm10
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm10[1,2],ymm3[3],ymm10[4],ymm3[5],ymm10[6,7],ymm3[8],ymm10[9,10],ymm3[11],ymm10[12],ymm3[13],ymm10[14,15]
-; AVX512F-SLOW-NEXT: vmovdqa %ymm3, %ymm11
+; AVX512F-SLOW-NEXT: vmovdqa 96(%rdi), %ymm9
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4],ymm3[5],ymm9[6,7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12],ymm3[13],ymm9[14,15]
+; AVX512F-SLOW-NEXT: vmovdqa %ymm3, %ymm10
; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5],ymm2[6],ymm3[7]
; AVX512F-SLOW-NEXT: vpshufb %ymm0, %ymm2, %ymm0
@@ -6729,21 +6744,21 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vpshufb %ymm1, %ymm2, %ymm1
; AVX512F-SLOW-NEXT: vpor %ymm0, %ymm1, %ymm0
; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5],ymm7[6],ymm6[7,8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13],ymm7[14],ymm6[15]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm11[0],ymm6[1],ymm11[2,3],ymm6[4],ymm11[5],ymm6[6],ymm11[7,8],ymm6[9],ymm11[10,11],ymm6[12],ymm11[13],ymm6[14],ymm11[15]
; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm0 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,8,9,18,19,28,29,22,23,16,17,26,27,128,128,128,128,128,128]
; AVX512F-SLOW-NEXT: vpshufb %ymm0, %ymm1, %ymm2
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm9[0],ymm8[1],ymm9[2],ymm8[3],ymm9[4,5],ymm8[6],ymm9[7,8],ymm8[9],ymm9[10],ymm8[11],ymm9[12,13],ymm8[14],ymm9[15]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm8[0],ymm7[1],ymm8[2],ymm7[3],ymm8[4,5],ymm7[6],ymm8[7,8],ymm7[9],ymm8[10],ymm7[11],ymm8[12,13],ymm7[14],ymm8[15]
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm3
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0,1],xmm3[2,3],xmm1[4,5,6],xmm3[7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = <2,3,12,13,6,7,0,1,10,11,4,5,14,15,128,128,128,128,128,128,128,128,128,128,128,128,u,u,u,u,u,u>
; AVX512F-SLOW-NEXT: vpshufb %ymm1, %ymm3, %ymm3
; AVX512F-SLOW-NEXT: vpor %ymm2, %ymm3, %ymm2
; AVX512F-SLOW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm10[0],ymm11[1],ymm10[2,3],ymm11[4],ymm10[5],ymm11[6],ymm10[7,8],ymm11[9],ymm10[10,11],ymm11[12],ymm10[13],ymm11[14],ymm10[15]
-; AVX512F-SLOW-NEXT: vmovdqu %ymm11, (%rsp) # 32-byte Spill
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm9[0],ymm10[1],ymm9[2,3],ymm10[4],ymm9[5],ymm10[6],ymm9[7,8],ymm10[9],ymm9[10,11],ymm10[12],ymm9[13],ymm10[14],ymm9[15]
; AVX512F-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-SLOW-NEXT: vmovdqu %ymm9, (%rsp) # 32-byte Spill
; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5],ymm2[6,7]
; AVX512F-SLOW-NEXT: vpshufb %ymm0, %ymm2, %ymm0
@@ -6753,39 +6768,40 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6],xmm3[7]
; AVX512F-SLOW-NEXT: vpshufb %ymm1, %ymm2, %ymm1
-; AVX512F-SLOW-NEXT: vpor %ymm0, %ymm1, %ymm0
-; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm15[0,1],ymm12[2],ymm15[3],ymm12[4],ymm15[5,6],ymm12[7],ymm15[8,9],ymm12[10],ymm15[11],ymm12[12],ymm15[13,14],ymm12[15]
+; AVX512F-SLOW-NEXT: vporq %ymm0, %ymm1, %ymm28
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm6 = [6,7,0,1,10,11,4,5,14,15,14,15,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm13[0,1],ymm15[2],ymm13[3],ymm15[4],ymm13[5,6],ymm15[7],ymm13[8,9],ymm15[10],ymm13[11],ymm15[12],ymm13[13,14],ymm15[15]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm15, %ymm18
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm13, %ymm24
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[6,7,6,7,8,9,4,5,14,15,8,9,2,3,12,13,22,23,22,23,24,25,20,21,30,31,24,25,18,19,28,29]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm13[0],xmm14[1],xmm13[2,3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm14, %xmm26
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm13, %xmm24
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[6,7,0,1,10,11,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm12[0],xmm14[1],xmm12[2,3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm12, %xmm16
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm14, %xmm30
+; AVX512F-SLOW-NEXT: vpshufb %xmm6, %xmm1, %xmm1
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm0[3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT: vmovdqa 144(%rdi), %xmm2
; AVX512F-SLOW-NEXT: vmovdqa 128(%rdi), %xmm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm3[2],xmm2[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm3, %xmm18
-; AVX512F-SLOW-NEXT: vmovdqa %xmm2, %xmm6
+; AVX512F-SLOW-NEXT: vmovdqa %xmm3, %xmm7
+; AVX512F-SLOW-NEXT: vmovdqa %xmm2, %xmm11
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm10[0],ymm11[1],ymm10[2],ymm11[3],ymm10[4,5],ymm11[6],ymm10[7,8],ymm11[9],ymm10[10],ymm11[11],ymm10[12,13],ymm11[14],ymm10[15]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm9[0],ymm10[1],ymm9[2],ymm10[3],ymm9[4,5],ymm10[6],ymm9[7,8],ymm10[9],ymm9[10],ymm10[11],ymm9[12,13],ymm10[14],ymm9[15]
; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4],ymm1[5],ymm2[6],ymm1[7]
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm5[1,2],ymm4[3],ymm5[4],ymm4[5],ymm5[6,7],ymm4[8],ymm5[9,10],ymm4[11],ymm5[12],ymm4[13],ymm5[14,15]
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[6,7,0,1,10,11,4,5,14,15,14,15,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29]
+; AVX512F-SLOW-NEXT: vpshufb %ymm6, %ymm1, %ymm1
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[6,7,0,1,10,11,4,5,14,15,8,9,12,13,14,15]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa 256(%rdi), %ymm3
-; AVX512F-SLOW-NEXT: vmovdqa 288(%rdi), %ymm4
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5],ymm4[6],ymm3[7,8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13],ymm4[14],ymm3[15]
-; AVX512F-SLOW-NEXT: vmovdqa %ymm3, %ymm5
+; AVX512F-SLOW-NEXT: vmovdqa 256(%rdi), %ymm12
+; AVX512F-SLOW-NEXT: vmovdqa 288(%rdi), %ymm15
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm12[0],ymm15[1],ymm12[2,3],ymm15[4],ymm12[5],ymm15[6],ymm12[7,8],ymm15[9],ymm12[10,11],ymm15[12],ymm12[13],ymm15[14],ymm12[15]
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1,2,3],xmm2[4,5],xmm3[6,7]
; AVX512F-SLOW-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
@@ -6796,294 +6812,294 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vmovdqa 464(%rdi), %xmm13
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm13[3,1,2,3]
+; AVX512F-SLOW-NEXT: vmovdqa 464(%rdi), %xmm8
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm8[3,1,2,3]
; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
; AVX512F-SLOW-NEXT: vmovdqa 448(%rdi), %xmm3
; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm3[0,2,2,3]
-; AVX512F-SLOW-NEXT: vmovdqa %xmm3, %xmm7
; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX512F-SLOW-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vmovdqa %xmm6, %xmm3
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm6[3,1,2,3]
+; AVX512F-SLOW-NEXT: vmovdqa %xmm11, %xmm6
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm11[3,1,2,3]
; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm18[0,2,2,3]
+; AVX512F-SLOW-NEXT: vmovdqa %xmm7, %xmm9
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm7[0,2,2,3]
; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX512F-SLOW-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm7[0],xmm13[1],xmm7[2,3]
-; AVX512F-SLOW-NEXT: vmovdqa %xmm7, %xmm10
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm3[0],xmm8[1],xmm3[2,3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm3, %xmm22
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = [6,7,2,3,4,5,6,7,6,7,2,3,12,13,6,7]
; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm20
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm7
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm23
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm10
; AVX512F-SLOW-NEXT: vmovdqa 576(%rdi), %ymm1
; AVX512F-SLOW-NEXT: vmovdqa 608(%rdi), %ymm2
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1],ymm1[2],ymm2[3],ymm1[4],ymm2[5,6],ymm1[7],ymm2[8,9],ymm1[10],ymm2[11],ymm1[12],ymm2[13,14],ymm1[15]
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm2, %ymm17
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm16
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm2, %ymm20
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm17
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm8 = xmm0[0,1,2],xmm1[3,4],xmm0[5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa 512(%rdi), %ymm6
-; AVX512F-SLOW-NEXT: vmovdqa 544(%rdi), %ymm2
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm6[1],ymm2[2,3],ymm6[4],ymm2[5],ymm6[6],ymm2[7,8],ymm6[9],ymm2[10,11],ymm6[12],ymm2[13],ymm6[14],ymm2[15]
-; AVX512F-SLOW-NEXT: vmovdqa %ymm2, %ymm11
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm11 = xmm0[0,1,2],xmm1[3,4],xmm0[5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa 512(%rdi), %ymm5
+; AVX512F-SLOW-NEXT: vmovdqa 544(%rdi), %ymm13
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm13[0],ymm5[1],ymm13[2,3],ymm5[4],ymm13[5],ymm5[6],ymm13[7,8],ymm5[9],ymm13[10,11],ymm5[12],ymm13[13],ymm5[14],ymm13[15]
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0],ymm0[1,2],ymm1[3],ymm0[4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 496(%rdi), %xmm27
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm27[3,1,2,3]
+; AVX512F-SLOW-NEXT: vmovdqa64 496(%rdi), %xmm21
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm21[3,1,2,3]
; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm1[2,1,2,3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 480(%rdi), %xmm28
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm14 = xmm28[0,2,2,3]
+; AVX512F-SLOW-NEXT: vmovdqa 480(%rdi), %xmm1
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm14 = xmm1[0,2,2,3]
+; AVX512F-SLOW-NEXT: vmovdqa %xmm1, %xmm7
; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm14 = xmm14[0,3,2,3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm14 = <u,u,u,u,u,u,u,u,8,9,2,3,12,13,6,7,16,17,26,27,26,27,30,31,24,25,18,19,28,29,22,23>
; AVX512F-SLOW-NEXT: vpshufb %ymm14, %ymm2, %ymm2
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = [12,13,14,15,4,5,14,15,8,9,2,3,12,13,6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm8, %xmm8
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm25 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
-; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm19, %zmm25, %zmm7
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm7, %zmm0
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm11, %xmm11
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm11[5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm27 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
+; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm19, %zmm27, %zmm10
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm10, %zmm0
; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0,1],ymm5[2],ymm4[3],ymm5[4],ymm4[5,6],ymm5[7],ymm4[8,9],ymm5[10],ymm4[11],ymm5[12],ymm4[13,14],ymm5[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm7
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm7[3,4],xmm0[5,6,7]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm15[0,1],ymm12[2],ymm15[3],ymm12[4],ymm15[5,6],ymm12[7],ymm15[8,9],ymm12[10],ymm15[11],ymm12[12],ymm15[13,14],ymm12[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm10
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm10[3,4],xmm0[5,6,7]
; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm0, %xmm0
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm15[0],ymm12[1],ymm15[2,3],ymm12[4],ymm15[5],ymm12[6],ymm15[7,8],ymm12[9],ymm15[10,11],ymm12[12],ymm15[13],ymm12[14],ymm15[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm7
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0],ymm2[1,2],ymm7[3],ymm2[4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm18, %ymm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm24, %ymm1
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5],ymm3[6],ymm1[7,8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13],ymm3[14],ymm1[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm10
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0],ymm2[1,2],ymm10[3],ymm2[4,5,6,7]
; AVX512F-SLOW-NEXT: vpshufb %ymm14, %ymm2, %ymm2
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm7 = xmm24[3,1,2,3]
-; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[2,1,2,3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm8 = xmm26[0,2,2,3]
-; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,3,2,3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm10 = xmm16[3,1,2,3]
+; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[2,1,2,3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm24
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm11 = xmm30[0,2,2,3]
+; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,3,2,3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1],ymm2[2,3,4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm18, %xmm1
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm1[0],xmm3[1],xmm1[2,3]
-; AVX512F-SLOW-NEXT: vmovdqa %xmm3, %xmm7
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm2, %xmm2
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm9[0],xmm6[1],xmm9[2,3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm9, %xmm25
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm4
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm2, %xmm2
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512F-SLOW-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm25, %zmm2 # 64-byte Folded Reload
+; AVX512F-SLOW-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm27, %zmm2 # 64-byte Folded Reload
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm0
; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1],xmm13[2],xmm10[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm10, %xmm29
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm13, %xmm30
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm0, %xmm0
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm2, %xmm31
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm6[0],ymm11[1],ymm6[2,3],ymm11[4],ymm6[5],ymm11[6],ymm6[7,8],ymm11[9],ymm6[10,11],ymm11[12],ymm6[13],ymm11[14],ymm6[15]
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm6, %ymm23
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm11, %ymm19
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm8
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm8[1],ymm2[2],ymm8[3],ymm2[4,5,6,7]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm8 = xmm28[0,3,2,3]
-; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[1,2,2,3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpsrlq $48, %xmm27, %xmm13
-; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm8[0],xmm13[0],xmm8[1],xmm13[1]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm4
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm4[0,1],xmm8[2],xmm4[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm2, %xmm2
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm5[0],ymm13[1],ymm5[2,3],ymm13[4],ymm5[5],ymm13[6],ymm5[7,8],ymm13[9],ymm5[10,11],ymm13[12],ymm5[13],ymm13[14],ymm5[15]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm13, %ymm26
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm5, %ymm22
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm10, %xmm11
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm11[1],ymm10[2],ymm11[3],ymm10[4,5,6,7]
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm11 = xmm7[0,3,2,3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm7, %xmm30
+; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,2,2,3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpsrlq $48, %xmm21, %xmm13
+; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm11[0],xmm13[0],xmm11[1],xmm13[1]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm13 = [2,3,12,13,0,1,0,1,10,11,4,5,14,15,8,9,18,19,28,29,16,17,16,17,26,27,20,21,30,31,24,25]
-; AVX512F-SLOW-NEXT: vpshufb %ymm13, %ymm2, %ymm2
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1,2],xmm2[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm16, %ymm6
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm17, %ymm3
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm8 = ymm6[0],ymm3[1,2],ymm6[3],ymm3[4],ymm6[5],ymm3[6,7],ymm6[8],ymm3[9,10],ymm6[11],ymm3[12],ymm6[13],ymm3[14,15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm8, %xmm14
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm8 = xmm14[0],xmm8[1],xmm14[2],xmm8[3]
+; AVX512F-SLOW-NEXT: vpshufb %ymm13, %ymm10, %ymm10
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2],xmm10[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm20, %ymm9
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm17, %ymm7
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm11 = ymm7[0],ymm9[1,2],ymm7[3],ymm9[4],ymm7[5],ymm9[6,7],ymm7[8],ymm9[9,10],ymm7[11],ymm9[12],ymm7[13],ymm9[14,15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm11, %xmm14
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm11 = xmm14[0],xmm11[1],xmm14[2],xmm11[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm14 = [0,1,2,3,6,7,0,1,10,11,4,5,14,15,8,9]
-; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm8, %xmm8
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm8[5,6,7]
-; AVX512F-SLOW-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm25, %zmm0 # 64-byte Folded Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm11, %xmm11
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm11[5,6,7]
+; AVX512F-SLOW-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm27, %zmm2 # 64-byte Folded Reload
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm10, %zmm2, %zmm2
+; AVX512F-SLOW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5],ymm1[6],ymm3[7,8],ymm1[9],ymm3[10,11],ymm1[12],ymm3[13],ymm1[14],ymm3[15]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm1, %ymm17
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm10
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm10[1],ymm2[2],ymm10[3],ymm2[4,5,6,7]
+; AVX512F-SLOW-NEXT: vpshufb %ymm13, %ymm2, %ymm2
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm10 = xmm24[0,3,2,3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm20
+; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[1,2,2,3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm16, %xmm1
+; AVX512F-SLOW-NEXT: vpsrlq $48, %xmm16, %xmm11
+; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm10[0],xmm11[0],xmm10[1],xmm11[1]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2],xmm2[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm12[0],ymm15[1,2],ymm12[3],ymm15[4],ymm12[5],ymm15[6,7],ymm12[8],ymm15[9,10],ymm12[11],ymm15[12],ymm12[13],ymm15[14,15]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm12, %ymm19
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm10, %xmm11
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm10 = xmm11[0],xmm10[1],xmm11[2],xmm10[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm10, %xmm10
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm10[5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm5
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm10 = xmm5[0,1],xmm6[2],xmm5[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm6, %xmm23
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm10, %xmm0
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm28, %zmm27, %zmm0
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm12[0],ymm15[1],ymm12[2,3],ymm15[4],ymm12[5],ymm15[6],ymm12[7,8],ymm15[9],ymm12[10,11],ymm15[12],ymm12[13],ymm15[14],ymm12[15]
-; AVX512F-SLOW-NEXT: vmovdqa %ymm15, %ymm10
-; AVX512F-SLOW-NEXT: vmovdqa %ymm12, %ymm9
-; AVX512F-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2],ymm2[3],ymm0[4,5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb %ymm13, %ymm0, %ymm0
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm26[0,3,2,3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm20
-; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpsrlq $48, %xmm24, %xmm8
-; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm0[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm5[0],ymm4[1,2],ymm5[3],ymm4[4],ymm5[5],ymm4[6,7],ymm5[8],ymm4[9,10],ymm5[11],ymm4[12],ymm5[13],ymm4[14,15]
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm5, %ymm18
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm4, %ymm26
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm8
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm8[0],xmm2[1],xmm8[2],xmm2[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm2, %xmm2
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm1[0,1],xmm7[2],xmm1[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm7, %xmm21
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm22
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm1
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm2, %xmm2
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512F-SLOW-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm25, %zmm2 # 64-byte Folded Reload
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm0
-; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm11[0],ymm5[1],ymm11[2,3],ymm5[4],ymm11[5],ymm5[6],ymm11[7,8],ymm5[9],ymm11[10,11],ymm5[12],ymm11[13],ymm5[14],ymm11[15]
+; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm18 # 32-byte Reload
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm18, %ymm0
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3],ymm12[4],ymm0[5],ymm12[6],ymm0[7,8],ymm12[9],ymm0[10,11],ymm12[12],ymm0[13],ymm12[14],ymm0[15]
; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6],ymm0[7]
-; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm17 # 32-byte Reload
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm2[6],ymm0[7]
+; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm24 # 32-byte Reload
; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm16 # 32-byte Reload
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm17, %ymm1
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm16, %ymm2
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0,1],ymm2[2],ymm1[3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8,9],ymm2[10],ymm1[11],ymm2[12],ymm1[13,14],ymm2[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm7
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm7[3,4],xmm2[5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = [4,5,14,15,8,9,2,3,12,13,10,11,0,1,10,11,20,21,30,31,24,25,18,19,28,29,26,27,16,17,26,27]
-; AVX512F-SLOW-NEXT: vpshufb %ymm7, %ymm0, %ymm8
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = [4,5,14,15,8,9,2,3,12,13,6,7,12,13,14,15]
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm2, %xmm2
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm8[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm8
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm12
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm13 = xmm12[0],xmm8[1],xmm12[2,3]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm8 = [0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
-; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm13, %xmm13
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm24, %ymm0
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm16, %ymm5
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm5[2],ymm0[3],ymm5[4],ymm0[5,6],ymm5[7],ymm0[8,9],ymm5[10],ymm0[11],ymm5[12],ymm0[13,14],ymm5[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm10
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm10 = xmm0[0,1,2],xmm10[3,4],xmm0[5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm0 = [4,5,14,15,8,9,2,3,12,13,10,11,0,1,10,11,20,21,30,31,24,25,18,19,28,29,26,27,16,17,26,27]
+; AVX512F-SLOW-NEXT: vpshufb %ymm0, %ymm2, %ymm2
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[4,5,14,15,8,9,2,3,12,13,6,7,12,13,14,15]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1,2],ymm2[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm13 = xmm8[0],xmm4[1],xmm8[2,3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm4, %xmm29
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm8, %xmm28
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
-; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm2, %zmm25, %zmm13
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm23, %ymm1
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm19, %ymm4
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm1[1],ymm4[2],ymm1[3],ymm4[4,5],ymm1[6],ymm4[7,8],ymm1[9],ymm4[10],ymm1[11],ymm4[12,13],ymm1[14],ymm4[15]
+; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm2, %zmm27, %zmm13
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm26, %ymm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm22, %ymm4
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm8[0],ymm4[1],ymm8[2],ymm4[3],ymm8[4,5],ymm4[6],ymm8[7,8],ymm4[9],ymm8[10],ymm4[11],ymm8[12,13],ymm4[14],ymm8[15]
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm14
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm14[1],ymm2[2,3,4,5,6,7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm14 = [4,5,14,15,12,13,2,3,12,13,6,7,0,1,10,11,20,21,30,31,28,29,18,19,28,29,22,23,16,17,26,27]
; AVX512F-SLOW-NEXT: vpshufb %ymm14, %ymm2, %ymm2
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm15 = xmm28[0,1,1,3]
-; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,7,6,7]
-; AVX512F-SLOW-NEXT: vpunpckhdq {{.*#+}} xmm15 = xmm15[2],xmm27[2],xmm15[3],xmm27[3]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm15 = xmm15[0,1,2],xmm2[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm15 = ymm3[0],ymm6[1],ymm3[2,3],ymm6[4],ymm3[5],ymm6[6],ymm3[7,8],ymm6[9],ymm3[10,11],ymm6[12],ymm3[13],ymm6[14],ymm3[15]
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm3, %ymm19
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm6, %ymm23
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm15, %xmm12
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0,1,2],xmm15[3,4],xmm12[5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm15 = [0,1,2,3,8,9,2,3,12,13,6,7,0,1,10,11]
-; AVX512F-SLOW-NEXT: vpshufb %xmm15, %xmm12, %xmm12
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm12[5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm6
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm11 = xmm30[0,1,1,3]
+; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,7,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm5
+; AVX512F-SLOW-NEXT: vpunpckhdq {{.*#+}} xmm11 = xmm11[2],xmm21[2],xmm11[3],xmm21[3]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2],xmm2[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm11 = ymm9[0],ymm7[1],ymm9[2,3],ymm7[4],ymm9[5],ymm7[6],ymm9[7,8],ymm7[9],ymm9[10,11],ymm7[12],ymm9[13],ymm7[14],ymm9[15]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm7, %ymm21
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm11, %xmm10
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2],xmm11[3,4],xmm10[5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm11 = [0,1,2,3,8,9,2,3,12,13,6,7,0,1,10,11]
+; AVX512F-SLOW-NEXT: vpshufb %xmm11, %xmm10, %xmm10
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm10[5,6,7]
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm2, %zmm13, %zmm2
; AVX512F-SLOW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm10[0],ymm9[1],ymm10[2],ymm9[3],ymm10[4,5],ymm9[6],ymm10[7,8],ymm9[9],ymm10[10],ymm9[11],ymm10[12,13],ymm9[14],ymm10[15]
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm10, %ymm31
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm12
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm12[1],ymm2[2,3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm17, %ymm2
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[3],ymm2[4,5],ymm3[6],ymm2[7,8],ymm3[9],ymm2[10],ymm3[11],ymm2[12,13],ymm3[14],ymm2[15]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm3, %ymm22
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm17, %ymm26
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm10
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm10[1],ymm2[2,3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpshufb %ymm14, %ymm2, %ymm2
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm12 = xmm20[0,1,1,3]
-; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,7,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm10
-; AVX512F-SLOW-NEXT: vpunpckhdq {{.*#+}} xmm12 = xmm12[2],xmm24[2],xmm12[3],xmm24[3]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0,1,2],xmm2[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX512F-SLOW-NEXT: vpblendw $173, (%rsp), %ymm3, %ymm12 # 32-byte Folded Reload
-; AVX512F-SLOW-NEXT: # ymm12 = mem[0],ymm3[1],mem[2,3],ymm3[4],mem[5],ymm3[6],mem[7,8],ymm3[9],mem[10,11],ymm3[12],mem[13],ymm3[14],mem[15]
-; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm13 = ymm12[2,3,0,1]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm13[6],ymm12[7]
-; AVX512F-SLOW-NEXT: vpshufb %ymm7, %ymm12, %ymm7
-; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX512F-SLOW-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm12 # 32-byte Folded Reload
-; AVX512F-SLOW-NEXT: # ymm12 = ymm3[0,1],mem[2],ymm3[3],mem[4],ymm3[5,6],mem[7],ymm3[8,9],mem[10],ymm3[11],mem[12],ymm3[13,14],mem[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm12, %xmm13
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0,1,2],xmm13[3,4],xmm12[5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm12, %xmm0
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm7[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm6
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm7 = xmm3[0],xmm6[1],xmm3[2,3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm7, %xmm7
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm0, %zmm25, %zmm7
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm18, %ymm0
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm26, %ymm3
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3],ymm0[4],ymm3[5],ymm0[6],ymm3[7,8],ymm0[9],ymm3[10,11],ymm0[12],ymm3[13],ymm0[14],ymm3[15]
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm18, %ymm24
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm8
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm8[0,1,2],xmm0[3,4],xmm8[5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm15, %xmm0, %xmm0
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm7, %zmm18
-; AVX512F-SLOW-NEXT: vmovdqa %ymm5, %ymm12
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0],ymm11[1],ymm5[2],ymm11[3],ymm5[4,5],ymm11[6],ymm5[7,8],ymm11[9],ymm5[10],ymm11[11],ymm5[12,13],ymm11[14],ymm5[15]
-; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4],ymm0[5],ymm2[6],ymm0[7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[6,7,0,1,10,11,4,5,14,15,14,15,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29]
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm10 = xmm20[0,1,1,3]
+; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,7,6,7]
+; AVX512F-SLOW-NEXT: vpunpckhdq {{.*#+}} xmm10 = xmm10[2],xmm1[2],xmm10[3],xmm1[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm17
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2],xmm2[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm19, %ymm1
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm15[0],ymm1[1],ymm15[2,3],ymm1[4],ymm15[5],ymm1[6],ymm15[7,8],ymm1[9],ymm15[10,11],ymm1[12],ymm15[13],ymm1[14],ymm15[15]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm15, %ymm31
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm10, %xmm13
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm10 = xmm13[0,1,2],xmm10[3,4],xmm13[5,6,7]
+; AVX512F-SLOW-NEXT: vpshufb %xmm11, %xmm10, %xmm10
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm10[5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX512F-SLOW-NEXT: vpblendw $82, (%rsp), %ymm1, %ymm10 # 32-byte Folded Reload
+; AVX512F-SLOW-NEXT: # ymm10 = ymm1[0],mem[1],ymm1[2,3],mem[4],ymm1[5],mem[6],ymm1[7,8],mem[9],ymm1[10,11],mem[12],ymm1[13],mem[14],ymm1[15]
+; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6],ymm10[7]
+; AVX512F-SLOW-NEXT: vpshufb %ymm0, %ymm10, %ymm0
+; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX512F-SLOW-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
+; AVX512F-SLOW-NEXT: # ymm10 = ymm1[0,1],mem[2],ymm1[3],mem[4],ymm1[5,6],mem[7],ymm1[8,9],mem[10],ymm1[11],mem[12],ymm1[13,14],mem[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm10, %xmm11
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2],xmm11[3,4],xmm10[5,6,7]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[4,5,14,15,8,9,2,3,12,13,6,7,12,13,14,15]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm3
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm10 = xmm3[0],xmm1[1],xmm3[2,3]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
+; AVX512F-SLOW-NEXT: vpternlogq $184, %zmm0, %zmm27, %zmm10
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm2, %zmm10, %zmm23
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm18, %ymm13
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm12[0],ymm13[1],ymm12[2],ymm13[3],ymm12[4,5],ymm13[6],ymm12[7,8],ymm13[9],ymm12[10],ymm13[11],ymm12[12,13],ymm13[14],ymm12[15]
+; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm0[2,3,0,1]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm10[4],ymm0[5],ymm10[6],ymm0[7]
; AVX512F-SLOW-NEXT: vmovdqa64 %ymm16, %ymm14
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm17, %ymm15
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm14[0],ymm15[1,2],ymm14[3],ymm15[4],ymm14[5],ymm15[6,7],ymm14[8],ymm15[9,10],ymm14[11],ymm15[12],ymm14[13],ymm15[14,15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm7
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm7[0],xmm2[1],xmm7[2],xmm2[3]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[6,7,0,1,10,11,4,5,14,15,8,9,12,13,14,15]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa %ymm1, %ymm3
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0,1],ymm1[2],ymm4[3],ymm1[4],ymm4[5,6],ymm1[7],ymm4[8,9],ymm1[10],ymm4[11],ymm1[12],ymm4[13,14],ymm1[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm7
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[6,7,6,7,8,9,4,5,14,15,8,9,2,3,12,13,22,23,22,23,24,25,20,21,30,31,24,25,18,19,28,29]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm27, %xmm6
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm1
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm7 = xmm6[0],xmm1[1],xmm6[2,3]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[6,7,0,1,10,11,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2],xmm2[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm24, %ymm15
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm14[0],ymm15[1,2],ymm14[3],ymm15[4],ymm14[5],ymm15[6,7],ymm14[8],ymm15[9,10],ymm14[11],ymm15[12],ymm14[13],ymm15[14,15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm10, %xmm11
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm10 = xmm11[0],xmm10[1],xmm11[2],xmm10[3]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[6,7,0,1,10,11,4,5,14,15,8,9,12,13,14,15]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = [6,7,0,1,10,11,4,5,14,15,14,15,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29]
+; AVX512F-SLOW-NEXT: vpshufb %ymm1, %ymm0, %ymm0
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa %ymm4, %ymm3
+; AVX512F-SLOW-NEXT: vmovdqa %ymm8, %ymm4
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm8[0,1],ymm3[2],ymm8[3],ymm3[4],ymm8[5,6],ymm3[7],ymm8[8,9],ymm3[10],ymm8[11],ymm3[12],ymm8[13,14],ymm3[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm10, %xmm11
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm10 = ymm10[6,7,6,7,8,9,4,5,14,15,8,9,2,3,12,13,22,23,22,23,24,25,20,21,30,31,24,25,18,19,28,29]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm11 = xmm5[0],xmm6[1],xmm5[2,3]
; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm7
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1],xmm5[2],xmm7[3]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm2, %zmm7, %zmm2
-; AVX512F-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm2
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm19, %ymm9
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm23, %ymm8
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5],ymm9[6],ymm8[7,8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13],ymm9[14],ymm8[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm7
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm7[1,2,3],xmm0[4,5],xmm7[6,7]
+; AVX512F-SLOW-NEXT: vmovdqa %xmm5, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm11, %xmm11
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2],xmm10[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm2
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm8 = xmm2[0,1],xmm1[2],xmm2[3]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm10, %zmm8, %zmm8
+; AVX512F-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm8
+; AVX512F-SLOW-NEXT: vmovdqa %ymm9, %ymm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm21, %ymm10
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm10[0],ymm9[1],ymm10[2,3],ymm9[4],ymm10[5],ymm9[6],ymm10[7,8],ymm9[9],ymm10[10,11],ymm9[12],ymm10[13],ymm9[14],ymm10[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm9
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm9[1,2,3],xmm0[4,5],xmm9[6,7]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,10,11,4,5,14,15,8,9,2,3,12,13]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm2, %ymm7
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm7[0],ymm0[1,2,3,4,5,6,7],ymm7[8],ymm0[9,10,11,12,13,14,15]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm21
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm12[0,1],ymm11[2],ymm12[3],ymm11[4],ymm12[5,6],ymm11[7],ymm12[8,9],ymm11[10],ymm12[11],ymm11[12],ymm12[13,14],ymm11[15]
-; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4],ymm0[5,6],ymm2[7]
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm15[0],ymm14[1],ymm15[2,3],ymm14[4],ymm15[5],ymm14[6],ymm15[7,8],ymm14[9],ymm15[10,11],ymm14[12],ymm15[13],ymm14[14],ymm15[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm7
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm11 = xmm7[0,1,2],xmm2[3,4],xmm7[5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = [8,9,2,3,12,13,6,7,4,5,6,7,4,5,14,15,24,25,18,19,28,29,22,23,20,21,22,23,20,21,30,31]
-; AVX512F-SLOW-NEXT: vpshufb %ymm2, %ymm0, %ymm0
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm7 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm7, %xmm11, %xmm11
+; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm8, %ymm9
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm9[0],ymm0[1,2,3,4,5,6,7],ymm9[8],ymm0[9,10,11,12,13,14,15]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm8, %zmm25
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm12[0,1],ymm13[2],ymm12[3],ymm13[4],ymm12[5,6],ymm13[7],ymm12[8,9],ymm13[10],ymm12[11],ymm13[12],ymm12[13,14],ymm13[15]
+; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm8 = ymm0[2,3,0,1]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm8[4],ymm0[5,6],ymm8[7]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm8 = ymm15[0],ymm14[1],ymm15[2,3],ymm14[4],ymm15[5],ymm14[6],ymm15[7,8],ymm14[9],ymm15[10,11],ymm14[12],ymm15[13],ymm14[14],ymm15[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm8, %xmm9
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm11 = xmm9[0,1,2],xmm8[3,4],xmm9[5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm9 = [8,9,2,3,12,13,6,7,4,5,6,7,4,5,14,15,24,25,18,19,28,29,22,23,20,21,22,23,20,21,30,31]
+; AVX512F-SLOW-NEXT: vpshufb %ymm9, %ymm0, %ymm0
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm8 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
+; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm11, %xmm11
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1,2],ymm0[3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm3[0],ymm4[1,2],ymm3[3],ymm4[4],ymm3[5],ymm4[6,7],ymm3[8],ymm4[9,10],ymm3[11],ymm4[12],ymm3[13],ymm4[14,15]
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2],ymm4[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm6[0,1],xmm1[2],xmm6[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm6[0,1],xmm7[2],xmm6[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = [8,9,14,15,4,5,6,7,0,1,10,11,4,5,14,15,24,25,30,31,20,21,22,23,16,17,26,27,20,21,30,31]
; AVX512F-SLOW-NEXT: vpshufb %ymm3, %ymm4, %ymm4
-; AVX512F-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm5 = [8,9,2,3,12,13,0,0,8,9,2,3,12,13,0,0]
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm1, %xmm1
+; AVX512F-SLOW-NEXT: vpshufb %xmm9, %xmm1, %xmm1
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 16-byte Folded Reload
@@ -7091,53 +7107,54 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: movb $7, %al
; AVX512F-SLOW-NEXT: kmovw %eax, %k1
; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm1 {%k1}
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm9[0],ymm8[1],ymm9[2],ymm8[3],ymm9[4,5],ymm8[6],ymm9[7,8],ymm8[9],ymm9[10],ymm8[11],ymm9[12,13],ymm8[14],ymm9[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm4
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm0[0,1],xmm4[2,3],xmm0[4,5,6],xmm4[7]
+; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm10[1],ymm2[2],ymm10[3],ymm2[4,5],ymm10[6],ymm2[7,8],ymm10[9],ymm2[10],ymm10[11],ymm2[12,13],ymm10[14],ymm2[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm5
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm0[0,1],xmm5[2,3],xmm0[4,5,6],xmm5[7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = [6,7,2,3,12,13,6,7,0,1,10,11,4,5,14,15]
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm4
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm1, %ymm6
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm6[0],ymm4[1,2,3,4,5,6,7],ymm6[8],ymm4[9,10,11,12,13,14,15]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm5
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0],ymm5[1,2,3,4,5,6,7],ymm4[8],ymm5[9,10,11,12,13,14,15]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm1, %zmm1
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm31, %ymm6
-; AVX512F-SLOW-NEXT: vpblendw $41, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm4 # 32-byte Folded Reload
-; AVX512F-SLOW-NEXT: # ymm4 = mem[0],ymm6[1,2],mem[3],ymm6[4],mem[5],ymm6[6,7],mem[8],ymm6[9,10],mem[11],ymm6[12],mem[13],ymm6[14,15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm6
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0],ymm4[1],ymm6[2],ymm4[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm22, %ymm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm26, %ymm4
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm2[0],ymm4[1,2],ymm2[3],ymm4[4],ymm2[5],ymm4[6,7],ymm2[8],ymm4[9,10],ymm2[11],ymm4[12],ymm2[13],ymm4[14,15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2],ymm4[3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpshufb %ymm3, %ymm4, %ymm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm4
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm10[0,1],xmm4[2],xmm10[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm4, %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm17, %xmm4
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm2[2],xmm4[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm9, %xmm4, %xmm4
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2],xmm3[3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm3, %zmm4, %zmm3
-; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX512F-SLOW-NEXT: vpblendw $148, (%rsp), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX512F-SLOW-NEXT: # ymm4 = ymm4[0,1],mem[2],ymm4[3],mem[4],ymm4[5,6],mem[7],ymm4[8,9],mem[10],ymm4[11],mem[12],ymm4[13,14],mem[15]
+; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX512F-SLOW-NEXT: vpblendw $107, (%rsp), %ymm2, %ymm4 # 32-byte Folded Reload
+; AVX512F-SLOW-NEXT: # ymm4 = mem[0,1],ymm2[2],mem[3],ymm2[4],mem[5,6],ymm2[7],mem[8,9],ymm2[10],mem[11],ymm2[12],mem[13,14],ymm2[15]
; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6],ymm5[7]
-; AVX512F-SLOW-NEXT: vpshufb %ymm2, %ymm4, %ymm2
-; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX512F-SLOW-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX512F-SLOW-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3],mem[4],ymm4[5],mem[6],ymm4[7,8],mem[9],ymm4[10,11],mem[12],ymm4[13],mem[14],ymm4[15]
+; AVX512F-SLOW-NEXT: vpshufb %ymm9, %ymm4, %ymm4
+; AVX512F-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX512F-SLOW-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm5 # 32-byte Folded Reload
+; AVX512F-SLOW-NEXT: # ymm5 = ymm2[0],mem[1],ymm2[2,3],mem[4],ymm2[5],mem[6],ymm2[7,8],mem[9],ymm2[10,11],mem[12],ymm2[13],mem[14],ymm2[15]
+; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3,4],xmm6[5,6,7]
+; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm5, %xmm5
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
+; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm4, %zmm0, %zmm3 {%k1}
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm19, %ymm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm31, %ymm4
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm2[1],ymm4[2],ymm2[3],ymm4[4,5],ymm2[6],ymm4[7,8],ymm2[9],ymm4[10],ymm2[11],ymm4[12,13],ymm2[14],ymm4[15]
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3,4],xmm5[5,6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm7, %xmm4, %xmm4
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3,4,5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm2, %zmm0, %zmm3 {%k1}
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm26, %ymm2
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm24, %ymm4
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[3],ymm2[4,5],ymm4[6],ymm2[7,8],ymm4[9],ymm2[10],ymm4[11],ymm2[12,13],ymm4[14],ymm2[15]
-; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm4
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm4[2,3],xmm2[4,5,6],xmm4[7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm2, %xmm0
-; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm3, %ymm2
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3],xmm4[4,5,6],xmm5[7]
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm0
+; AVX512F-SLOW-NEXT: vextracti64x4 $1, %zmm3, %ymm4
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2,3,4,5,6,7],ymm4[8],ymm0[9,10,11,12,13,14,15]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm3, %zmm0
; AVX512F-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-SLOW-NEXT: vmovaps %zmm2, (%rsi)
@@ -7149,19 +7166,19 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vmovaps %zmm2, (%rdx)
; AVX512F-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-SLOW-NEXT: vmovaps %zmm2, 64(%rcx)
-; AVX512F-SLOW-NEXT: vmovdqa64 %zmm18, (%rcx)
-; AVX512F-SLOW-NEXT: vmovdqa64 %zmm21, 64(%r8)
+; AVX512F-SLOW-NEXT: vmovdqa64 %zmm23, (%rcx)
+; AVX512F-SLOW-NEXT: vmovdqa64 %zmm25, 64(%r8)
; AVX512F-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-SLOW-NEXT: vmovaps %zmm2, (%r8)
; AVX512F-SLOW-NEXT: vmovdqa64 %zmm1, 64(%r9)
; AVX512F-SLOW-NEXT: vmovdqa64 %zmm0, (%r9)
-; AVX512F-SLOW-NEXT: addq $584, %rsp # imm = 0x248
+; AVX512F-SLOW-NEXT: addq $552, %rsp # imm = 0x228
; AVX512F-SLOW-NEXT: vzeroupper
; AVX512F-SLOW-NEXT: retq
;
; AVX512F-FAST-LABEL: load_i16_stride5_vf64:
; AVX512F-FAST: # %bb.0:
-; AVX512F-FAST-NEXT: subq $712, %rsp # imm = 0x2C8
+; AVX512F-FAST-NEXT: subq $552, %rsp # imm = 0x228
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = [4,5,14,15,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT: vmovdqa 496(%rdi), %xmm1
; AVX512F-FAST-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7178,20 +7195,21 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vmovdqa 576(%rdi), %ymm7
; AVX512F-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT: vmovdqa 608(%rdi), %ymm8
-; AVX512F-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa 352(%rdi), %ymm1
-; AVX512F-FAST-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa 320(%rdi), %ymm4
-; AVX512F-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm4[0],ymm1[1],ymm4[2,3],ymm1[4],ymm4[5],ymm1[6],ymm4[7,8],ymm1[9],ymm4[10,11],ymm1[12],ymm4[13],ymm1[14],ymm4[15]
+; AVX512F-FAST-NEXT: vmovdqu %ymm8, (%rsp) # 32-byte Spill
+; AVX512F-FAST-NEXT: vmovdqa 352(%rdi), %ymm4
+; AVX512F-FAST-NEXT: vmovdqa 320(%rdi), %ymm5
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5],ymm4[6],ymm5[7,8],ymm4[9],ymm5[10,11],ymm4[12],ymm5[13],ymm4[14],ymm5[15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm5, %ymm28
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm30
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm1, %xmm4
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm1[0],xmm4[1,2,3],xmm1[4,5],xmm4[6,7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <0,1,10,11,4,5,14,15,8,9,2,3,12,13,128,128,128,128,128,128,128,128,128,128,128,128,u,u,u,u,u,u>
; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm4
; AVX512F-FAST-NEXT: vmovdqa 384(%rdi), %ymm6
-; AVX512F-FAST-NEXT: vmovdqa 416(%rdi), %ymm5
-; AVX512F-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm6[0],ymm5[1,2],ymm6[3],ymm5[4],ymm6[5],ymm5[6,7],ymm6[8],ymm5[9,10],ymm6[11],ymm5[12],ymm6[13],ymm5[14,15]
+; AVX512F-FAST-NEXT: vmovdqa 416(%rdi), %ymm9
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm6[0],ymm9[1,2],ymm6[3],ymm9[4],ymm6[5],ymm9[6,7],ymm6[8],ymm9[9,10],ymm6[11],ymm9[12],ymm6[13],ymm9[14,15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm9, %ymm26
+; AVX512F-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT: vmovdqa64 %ymm6, %ymm27
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [1,3,0,2,4,6,1,3]
; AVX512F-FAST-NEXT: vpermd %ymm5, %ymm6, %ymm5
@@ -7213,46 +7231,46 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0,1,2,3,4],ymm5[5,6,7]
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm18 = [0,3,1,3,0,3,5,7]
-; AVX512F-FAST-NEXT: vmovdqa64 448(%rdi), %ymm29
-; AVX512F-FAST-NEXT: vpermd %ymm29, %ymm18, %ymm7
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm8 = [4,5,2,3,4,5,6,7,8,9,2,3,4,5,10,11,20,21,18,19,20,21,22,23,24,25,18,19,20,21,26,27]
-; AVX512F-FAST-NEXT: vpshufb %ymm8, %ymm7, %ymm7
-; AVX512F-FAST-NEXT: vmovdqa %ymm8, %ymm14
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm26 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
-; AVX512F-FAST-NEXT: vpternlogq $184, %zmm4, %zmm26, %zmm7
+; AVX512F-FAST-NEXT: vmovdqa 448(%rdi), %ymm7
+; AVX512F-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-FAST-NEXT: vpermd %ymm7, %ymm18, %ymm7
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = [4,5,2,3,4,5,6,7,8,9,2,3,4,5,10,11,20,21,18,19,20,21,22,23,24,25,18,19,20,21,26,27]
+; AVX512F-FAST-NEXT: vpshufb %ymm11, %ymm7, %ymm7
+; AVX512F-FAST-NEXT: vmovdqa %ymm11, %ymm14
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm8 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
+; AVX512F-FAST-NEXT: vpternlogq $184, %zmm4, %zmm8, %zmm7
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm8, %zmm16
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm7, %zmm4
; AVX512F-FAST-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vmovdqa 176(%rdi), %xmm4
; AVX512F-FAST-NEXT: vpshufb %xmm0, %xmm4, %xmm0
; AVX512F-FAST-NEXT: vmovdqa64 %xmm4, %xmm31
-; AVX512F-FAST-NEXT: vmovdqa 160(%rdi), %xmm4
-; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm3
-; AVX512F-FAST-NEXT: vmovdqa %xmm4, %xmm15
+; AVX512F-FAST-NEXT: vmovdqa 160(%rdi), %xmm15
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm15, %xmm3
; AVX512F-FAST-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; AVX512F-FAST-NEXT: vmovdqa 64(%rdi), %ymm7
-; AVX512F-FAST-NEXT: vmovdqa 96(%rdi), %ymm8
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm7[0],ymm8[1,2],ymm7[3],ymm8[4],ymm7[5],ymm8[6,7],ymm7[8],ymm8[9,10],ymm7[11],ymm8[12],ymm7[13],ymm8[14,15]
-; AVX512F-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-FAST-NEXT: vmovdqa 64(%rdi), %ymm11
+; AVX512F-FAST-NEXT: vmovdqa 96(%rdi), %ymm5
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm11[0],ymm5[1,2],ymm11[3],ymm5[4],ymm11[5],ymm5[6,7],ymm11[8],ymm5[9,10],ymm11[11],ymm5[12],ymm11[13],ymm5[14,15]
+; AVX512F-FAST-NEXT: vmovdqa %ymm5, %ymm7
; AVX512F-FAST-NEXT: vpermd %ymm3, %ymm6, %ymm3
; AVX512F-FAST-NEXT: vpshufb %ymm13, %ymm3, %ymm3
; AVX512F-FAST-NEXT: vmovdqa (%rdi), %ymm5
; AVX512F-FAST-NEXT: vmovdqa 32(%rdi), %ymm6
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm5[0],ymm6[1],ymm5[2,3],ymm6[4],ymm5[5],ymm6[6],ymm5[7,8],ymm6[9],ymm5[10,11],ymm6[12],ymm5[13],ymm6[14],ymm5[15]
-; AVX512F-FAST-NEXT: vmovdqa %ymm5, %ymm11
+; AVX512F-FAST-NEXT: vmovdqa %ymm5, %ymm12
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0],xmm5[1,2,3],xmm4[4,5],xmm5[6,7]
; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm1
; AVX512F-FAST-NEXT: vpor %ymm3, %ymm1, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa 192(%rdi), %ymm4
-; AVX512F-FAST-NEXT: vmovdqa 224(%rdi), %ymm5
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5],ymm4[6],ymm5[7,8],ymm4[9],ymm5[10,11],ymm4[12],ymm5[13],ymm4[14],ymm5[15]
-; AVX512F-FAST-NEXT: vmovdqa %ymm5, %ymm13
-; AVX512F-FAST-NEXT: vmovdqa %ymm4, %ymm5
+; AVX512F-FAST-NEXT: vmovdqa 192(%rdi), %ymm8
+; AVX512F-FAST-NEXT: vmovdqa 224(%rdi), %ymm13
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm13[0],ymm8[1],ymm13[2,3],ymm8[4],ymm13[5],ymm8[6],ymm13[7,8],ymm8[9],ymm13[10,11],ymm8[12],ymm13[13],ymm8[14],ymm13[15]
; AVX512F-FAST-NEXT: vpermd %ymm3, %ymm10, %ymm3
; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm3, %ymm3
-; AVX512F-FAST-NEXT: vmovdqa 256(%rdi), %ymm12
+; AVX512F-FAST-NEXT: vmovdqa 256(%rdi), %ymm5
; AVX512F-FAST-NEXT: vmovdqa 288(%rdi), %ymm9
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm9[0,1],ymm12[2],ymm9[3],ymm12[4],ymm9[5,6],ymm12[7],ymm9[8,9],ymm12[10],ymm9[11],ymm12[12],ymm9[13,14],ymm12[15]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm9[0,1],ymm5[2],ymm9[3],ymm5[4],ymm9[5,6],ymm5[7],ymm9[8,9],ymm5[10],ymm9[11],ymm5[12],ymm9[13,14],ymm5[15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm5, %ymm25
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm4, %xmm10
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2],xmm10[3,4],xmm4[5,6,7]
; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm4, %xmm2
@@ -7262,123 +7280,124 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vmovdqa64 128(%rdi), %ymm23
; AVX512F-FAST-NEXT: vpermd %ymm23, %ymm18, %ymm2
; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm2, %ymm2
-; AVX512F-FAST-NEXT: vpternlogq $184, %zmm1, %zmm26, %zmm2
+; AVX512F-FAST-NEXT: vpternlogq $184, %zmm1, %zmm16, %zmm2
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm0
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0],ymm11[1],ymm6[2],ymm11[3],ymm6[4,5],ymm11[6],ymm6[7,8],ymm11[9],ymm6[10],ymm11[11],ymm6[12,13],ymm11[14],ymm6[15]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm11, %ymm30
-; AVX512F-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm6, %ymm17
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0],ymm12[1],ymm6[2],ymm12[3],ymm6[4,5],ymm12[6],ymm6[7,8],ymm12[9],ymm6[10],ymm12[11],ymm6[12,13],ymm12[14],ymm6[15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm12, %ymm21
+; AVX512F-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm6, %ymm29
; AVX512F-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6],xmm1[7]
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm21 = <2,u,u,u,4,7,1,6>
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm8[0],ymm7[1],ymm8[2,3],ymm7[4],ymm8[5],ymm7[6],ymm8[7,8],ymm7[9],ymm8[10,11],ymm7[12],ymm8[13],ymm7[14],ymm8[15]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm8, %ymm22
-; AVX512F-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vpermd %ymm1, %ymm21, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm17 = <2,u,u,u,4,7,1,6>
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm7[0],ymm11[1],ymm7[2,3],ymm11[4],ymm7[5],ymm11[6],ymm7[7,8],ymm11[9],ymm7[10,11],ymm11[12],ymm7[13],ymm11[14],ymm7[15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm11, %ymm19
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm7, %ymm22
+; AVX512F-FAST-NEXT: vpermd %ymm1, %ymm17, %ymm1
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,0,1,18,19,20,21,26,27,16,17,30,31,128,128,128,128,128,128]
; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm1, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <2,3,12,13,6,7,0,1,10,11,4,5,14,15,128,128,128,128,128,128,128,128,128,128,128,128,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm0, %ymm0
-; AVX512F-FAST-NEXT: vporq %ymm1, %ymm0, %ymm18
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <2,3,12,13,6,7,0,1,10,11,4,5,14,15,128,128,128,128,128,128,128,128,128,128,128,128,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm0
+; AVX512F-FAST-NEXT: vpor %ymm1, %ymm0, %ymm10
; AVX512F-FAST-NEXT: vpsrlq $48, %xmm31, %xmm0
-; AVX512F-FAST-NEXT: vpbroadcastd {{.*#+}} xmm4 = [2,3,12,13,2,3,12,13,2,3,12,13,2,3,12,13]
-; AVX512F-FAST-NEXT: vpshufb %xmm4, %xmm15, %xmm1
-; AVX512F-FAST-NEXT: vmovdqa %xmm15, %xmm11
+; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm15, %xmm1
; AVX512F-FAST-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm20 = <0,2,5,7,4,7,u,u>
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0],ymm13[1],ymm5[2,3],ymm13[4],ymm5[5],ymm13[6],ymm5[7,8],ymm13[9],ymm5[10,11],ymm13[12],ymm5[13],ymm13[14],ymm5[15]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm13, %ymm19
-; AVX512F-FAST-NEXT: vpermd %ymm0, %ymm20, %ymm10
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [2,3,4,5,4,5,0,1,6,7,8,9,14,15,4,5,18,19,20,21,20,21,16,17,22,23,24,25,30,31,20,21]
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm10, %ymm10
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm10[3,4,5,6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm10[4,5,6,7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm13[1],ymm8[2,3],ymm13[4],ymm8[5],ymm13[6],ymm8[7,8],ymm13[9],ymm8[10,11],ymm13[12],ymm8[13],ymm13[14],ymm8[15]
+; AVX512F-FAST-NEXT: vmovdqa %ymm8, %ymm5
+; AVX512F-FAST-NEXT: vpermd %ymm0, %ymm20, %ymm4
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [2,3,4,5,4,5,0,1,6,7,8,9,14,15,4,5,18,19,20,21,20,21,16,17,22,23,24,25,30,31,20,21]
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm4, %ymm4
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3,4,5,6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX512F-FAST-NEXT: vbroadcasti32x4 {{.*#+}} ymm24 = [1,4,6,3,1,4,6,3]
; AVX512F-FAST-NEXT: # ymm24 = mem[0,1,2,3,0,1,2,3]
; AVX512F-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm25, %ymm12
+; AVX512F-FAST-NEXT: vmovdqu64 %ymm25, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm12[0],ymm9[1,2],ymm12[3],ymm9[4],ymm12[5],ymm9[6,7],ymm12[8],ymm9[9,10],ymm12[11],ymm9[12],ymm12[13],ymm9[14,15]
-; AVX512F-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT: vpermd %ymm1, %ymm24, %ymm0
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [0,1,2,3,2,3,4,5,10,11,0,1,14,15,8,9,16,17,18,19,18,19,20,21,26,27,16,17,30,31,24,25]
; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm0, %ymm0
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3,4],ymm0[5,6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6,7]
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm25 = [1,3,2,3,1,3,6,7]
-; AVX512F-FAST-NEXT: vpermd %ymm23, %ymm25, %ymm10
+; AVX512F-FAST-NEXT: vpermd %ymm23, %ymm25, %ymm4
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = [0,1,6,7,4,5,6,7,8,9,0,1,6,7,8,9,16,17,22,23,20,21,22,23,24,25,16,17,22,23,24,25]
-; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm10, %ymm10
-; AVX512F-FAST-NEXT: vpternlogq $184, %zmm18, %zmm26, %zmm10
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm10, %zmm0
+; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm4, %ymm4
+; AVX512F-FAST-NEXT: vpternlogq $184, %zmm10, %zmm16, %zmm4
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm16, %zmm10
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm4, %zmm0
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm18 # 32-byte Reload
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm28, %ymm18
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm28
; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm0
-; AVX512F-FAST-NEXT: vpblendw $181, (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: # ymm0 = mem[0],ymm0[1],mem[2],ymm0[3],mem[4,5],ymm0[6],mem[7,8],ymm0[9],mem[10],ymm0[11],mem[12,13],ymm0[14],mem[15]
-; AVX512F-FAST-NEXT: vextracti128 $1, %ymm0, %xmm10
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm10[2,3],xmm0[4,5,6],xmm10[7]
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm0, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm27, %ymm13
-; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm7[0],ymm13[1],ymm7[2,3],ymm13[4],ymm7[5],ymm13[6],ymm7[7,8],ymm13[9],ymm7[10,11],ymm13[12],ymm7[13],ymm13[14],ymm7[15]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm7, %ymm28
-; AVX512F-FAST-NEXT: vpermd %ymm3, %ymm21, %ymm3
-; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX512F-FAST-NEXT: vpor %ymm3, %ymm0, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm4
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1],ymm4[2],ymm0[3],ymm4[4,5],ymm0[6],ymm4[7,8],ymm0[9],ymm4[10],ymm0[11],ymm4[12,13],ymm0[14],ymm4[15]
+; AVX512F-FAST-NEXT: vextracti128 $1, %ymm0, %xmm4
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3],xmm0[4,5,6],xmm4[7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm27, %ymm7
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm26, %ymm4
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm7[1],ymm4[2,3],ymm7[4],ymm4[5],ymm7[6],ymm4[7,8],ymm7[9],ymm4[10,11],ymm7[12],ymm4[13],ymm7[14],ymm4[15]
+; AVX512F-FAST-NEXT: vpermd %ymm4, %ymm17, %ymm4
+; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm4, %ymm4
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm0
+; AVX512F-FAST-NEXT: vpor %ymm4, %ymm0, %ymm0
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm16 # 32-byte Reload
-; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm16, %ymm3
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm15[0],ymm3[1],ymm15[2,3],ymm3[4],ymm15[5],ymm3[6],ymm15[7,8],ymm3[9],ymm15[10,11],ymm3[12],ymm15[13],ymm3[14],ymm15[15]
-; AVX512F-FAST-NEXT: vpermd %ymm3, %ymm20, %ymm3
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX512F-FAST-NEXT: vpshufb %xmm4, %xmm7, %xmm3
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm16, %ymm4
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm11[0],ymm4[1],ymm11[2,3],ymm4[4],ymm11[5],ymm4[6],ymm11[7,8],ymm4[9],ymm11[10,11],ymm4[12],ymm11[13],ymm4[14],ymm11[15]
+; AVX512F-FAST-NEXT: vpermd %ymm4, %ymm20, %ymm4
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm6, %xmm2
; AVX512F-FAST-NEXT: vmovdqa64 {{[-0-9]+}}(%r{{[sb]}}p), %xmm27 # 16-byte Reload
; AVX512F-FAST-NEXT: vpsrlq $48, %xmm27, %xmm4
-; AVX512F-FAST-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm2[3,4,5,6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm6[0],ymm8[1,2],ymm6[3],ymm8[4],ymm6[5],ymm8[6,7],ymm6[8],ymm8[9,10],ymm6[11],ymm8[12],ymm6[13],ymm8[14,15]
+; AVX512F-FAST-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3,4,5,6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-FAST-NEXT: vmovdqu (%rsp), %ymm8 # 32-byte Reload
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0],ymm8[1,2],ymm4[3],ymm8[4],ymm4[5],ymm8[6,7],ymm4[8],ymm8[9,10],ymm4[11],ymm8[12],ymm4[13],ymm8[14,15]
; AVX512F-FAST-NEXT: vpermd %ymm3, %ymm24, %ymm3
; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm3, %ymm1
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
-; AVX512F-FAST-NEXT: vpermd %ymm29, %ymm25, %ymm2
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm17 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpermd %ymm17, %ymm25, %ymm2
; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm2, %ymm2
-; AVX512F-FAST-NEXT: vpternlogq $184, %zmm0, %zmm26, %zmm2
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm26, %zmm21
+; AVX512F-FAST-NEXT: vpternlogq $184, %zmm0, %zmm10, %zmm2
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm0
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm9[0],ymm12[1],ymm9[2,3],ymm12[4],ymm9[5],ymm12[6],ymm9[7,8],ymm12[9],ymm9[10,11],ymm12[12],ymm9[13],ymm12[14],ymm9[15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm0[3,4],xmm1[5,6,7]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm11[u,u,u,u,u,u,u,u,4,5,14,15,u,u,u,u]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm11, %xmm20
-; AVX512F-FAST-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm0[2],xmm31[2],xmm0[3],xmm31[3]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm15[u,u,u,u,u,u,u,u,4,5,14,15,u,u,u,u]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm15, %xmm20
+; AVX512F-FAST-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm31[2],xmm0[3],xmm31[3]
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm24 = <0,3,5,2,5,7,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, %ymm12
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm12[0],ymm5[1],ymm12[2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7,8],ymm5[9],ymm12[10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm5, %ymm26
-; AVX512F-FAST-NEXT: vpermd %ymm0, %ymm24, %ymm4
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [0,1,6,7,2,3,2,3,4,5,10,11,0,1,14,15,16,17,22,23,18,19,18,19,20,21,26,27,16,17,30,31]
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm4, %ymm4
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3,4,5,6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,2,3,8,9,2,3,12,13,6,7,0,1,10,11]
-; AVX512F-FAST-NEXT: vpshufb %xmm4, %xmm1, %xmm1
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm13[0],ymm5[1],ymm13[2],ymm5[3],ymm13[4,5],ymm5[6],ymm13[7,8],ymm5[9],ymm13[10],ymm5[11],ymm13[12,13],ymm5[14],ymm13[15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm5, %ymm30
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm13, %ymm26
+; AVX512F-FAST-NEXT: vpermd %ymm2, %ymm24, %ymm3
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm12 = [0,1,6,7,2,3,2,3,4,5,10,11,0,1,14,15,16,17,22,23,18,19,18,19,20,21,26,27,16,17,30,31]
+; AVX512F-FAST-NEXT: vpshufb %ymm12, %ymm3, %ymm3
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3,4,5,6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,8,9,2,3,12,13,6,7,0,1,10,11]
+; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm1, %xmm1
; AVX512F-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm3[0,1,2,3,4],ymm1[5,6,7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm17, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm29, %ymm1
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm0[2],ymm1[3],ymm0[4],ymm1[5,6],ymm0[7],ymm1[8,9],ymm0[10],ymm1[11],ymm0[12],ymm1[13,14],ymm0[15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm1, %xmm3
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm5 = xmm1[0,1,2],xmm3[3,4],xmm1[5,6,7]
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm25 = <0,2,u,u,5,7,2,4>
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm19 # 32-byte Reload
; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, %ymm0
; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, %ymm1
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5],ymm1[6],ymm0[7,8],ymm1[9],ymm0[10,11],ymm1[12],ymm0[13],ymm1[14],ymm0[15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, %ymm29
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, %ymm21
; AVX512F-FAST-NEXT: vpermd %ymm1, %ymm25, %ymm3
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [0,1,6,7,8,9,14,15,4,5,6,7,0,1,6,7,16,17,22,23,24,25,30,31,20,21,22,23,16,17,22,23]
; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm3, %ymm0
@@ -7388,56 +7407,55 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [1,4,6,0,1,4,6,0]
; AVX512F-FAST-NEXT: # ymm5 = mem[0,1,0,1]
; AVX512F-FAST-NEXT: vpermd %ymm23, %ymm5, %ymm14
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = [4,5,2,3,4,5,6,7,8,9,2,3,4,5,10,11,20,21,18,19,20,21,22,23,24,25,18,19,20,21,26,27]
-; AVX512F-FAST-NEXT: vpshufb %ymm11, %ymm14, %ymm14
-; AVX512F-FAST-NEXT: vpternlogq $184, %zmm0, %zmm21, %zmm14
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = [4,5,2,3,4,5,6,7,8,9,2,3,4,5,10,11,20,21,18,19,20,21,22,23,24,25,18,19,20,21,26,27]
+; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm14, %ymm14
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm15 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
+; AVX512F-FAST-NEXT: vpternlogq $184, %zmm0, %zmm15, %zmm14
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm10, %zmm14, %zmm0
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm6[1],ymm8[2,3],ymm6[4],ymm8[5],ymm6[6],ymm8[7,8],ymm6[9],ymm8[10,11],ymm6[12],ymm8[13],ymm6[14],ymm8[15]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm4[1],ymm8[2,3],ymm4[4],ymm8[5],ymm4[6],ymm8[7,8],ymm4[9],ymm8[10,11],ymm4[12],ymm8[13],ymm4[14],ymm8[15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm0, %xmm10
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm10[0,1,2],xmm0[3,4],xmm10[5,6,7]
-; AVX512F-FAST-NEXT: vpshufb %xmm4, %xmm0, %xmm0
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm16, %ymm6
-; AVX512F-FAST-NEXT: vmovdqa %ymm15, %ymm8
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm6[0],ymm15[1],ymm6[2],ymm15[3],ymm6[4,5],ymm15[6],ymm6[7,8],ymm15[9],ymm6[10],ymm15[11],ymm6[12,13],ymm15[14],ymm6[15]
-; AVX512F-FAST-NEXT: vpermd %ymm4, %ymm24, %ymm4
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm4, %ymm2
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm4 = xmm7[u,u,u,u,u,u,u,u,4,5,14,15,u,u,u,u]
-; AVX512F-FAST-NEXT: vmovdqa %xmm7, %xmm14
+; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm16, %ymm8
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm8[0],ymm11[1],ymm8[2],ymm11[3],ymm8[4,5],ymm11[6],ymm8[7,8],ymm11[9],ymm8[10],ymm11[11],ymm8[12,13],ymm11[14],ymm8[15]
+; AVX512F-FAST-NEXT: vmovdqa %ymm11, %ymm13
+; AVX512F-FAST-NEXT: vpermd %ymm2, %ymm24, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm12, %ymm2, %ymm2
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm4 = xmm6[u,u,u,u,u,u,u,u,4,5,14,15,u,u,u,u]
+; AVX512F-FAST-NEXT: vmovdqa %xmm6, %xmm14
; AVX512F-FAST-NEXT: vpunpckhdq {{.*#+}} xmm4 = xmm4[2],xmm27[2],xmm4[3],xmm27[3]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm27, %xmm15
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm27, %xmm6
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2],xmm2[3,4,5,6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm4
-; AVX512F-FAST-NEXT: vmovdqu (%rsp), %ymm7 # 32-byte Reload
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm7[0,1],ymm4[2],ymm7[3],ymm4[4],ymm7[5,6],ymm4[7],ymm7[8,9],ymm4[10],ymm7[11],ymm4[12],ymm7[13,14],ymm4[15]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm7, %ymm30
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm22
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm11
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm28, %ymm12
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm12[0,1],ymm11[2],ymm12[3],ymm11[4],ymm12[5,6],ymm11[7],ymm12[8,9],ymm11[10],ymm12[11],ymm11[12],ymm12[13,14],ymm11[15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm2, %xmm4
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm4[3,4],xmm2[5,6,7]
; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm2, %xmm2
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm28, %ymm4
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm13[0],ymm4[1],ymm13[2,3],ymm4[4],ymm13[5],ymm4[6],ymm13[7,8],ymm4[9],ymm13[10,11],ymm4[12],ymm13[13],ymm4[14],ymm13[15]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm13, %ymm18
+; AVX512F-FAST-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm3 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: # ymm3 = ymm7[0],mem[1],ymm7[2,3],mem[4],ymm7[5],mem[6],ymm7[7,8],mem[9],ymm7[10,11],mem[12],ymm7[13],mem[14],ymm7[15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm7, %ymm16
; AVX512F-FAST-NEXT: vpermd %ymm3, %ymm25, %ymm3
; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm3, %ymm1
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
-; AVX512F-FAST-NEXT: vpermd %ymm29, %ymm5, %ymm2
-; AVX512F-FAST-NEXT: vpshufb %ymm11, %ymm2, %ymm2
-; AVX512F-FAST-NEXT: vpternlogq $184, %zmm1, %zmm21, %zmm2
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm0
-; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm31, %xmm7
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm20, %xmm9
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm7[0],xmm9[1],xmm7[2,3]
+; AVX512F-FAST-NEXT: vpermd %ymm17, %ymm5, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm2, %ymm2
+; AVX512F-FAST-NEXT: vpternlogq $184, %zmm1, %zmm15, %zmm2
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm28
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm31, %xmm15
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm20, %xmm7
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm15[0],xmm7[1],xmm15[2,3]
; AVX512F-FAST-NEXT: vpbroadcastq {{.*#+}} xmm0 = [6,7,0,1,10,11,0,0,6,7,0,1,10,11,0,0]
; AVX512F-FAST-NEXT: vpshufb %xmm0, %xmm1, %xmm2
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <1,3,6,0,5,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm18 = <1,3,6,0,5,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm9
; AVX512F-FAST-NEXT: vmovdqa64 %ymm26, %ymm10
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm12[0,1],ymm10[2],ymm12[3],ymm10[4],ymm12[5,6],ymm10[7],ymm12[8,9],ymm10[10],ymm12[11],ymm10[12],ymm12[13,14],ymm10[15]
-; AVX512F-FAST-NEXT: vpermd %ymm1, %ymm3, %ymm4
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm10[0,1],ymm9[2],ymm10[3],ymm9[4],ymm10[5,6],ymm9[7],ymm10[8,9],ymm9[10],ymm10[11],ymm9[12],ymm10[13,14],ymm9[15]
+; AVX512F-FAST-NEXT: vpermd %ymm1, %ymm18, %ymm4
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [2,3,2,3,4,5,0,1,6,7,8,9,14,15,4,5,18,19,18,19,20,21,16,17,22,23,24,25,30,31,20,21]
; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm4
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm4[3,4,5,6,7]
@@ -7445,24 +7463,25 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm4 = [2,4,7,0,2,4,7,0]
; AVX512F-FAST-NEXT: # ymm4 = mem[0,1,0,1]
; AVX512F-FAST-NEXT: vpermd %ymm23, %ymm4, %ymm5
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = [0,1,6,7,4,5,6,7,8,9,0,1,6,7,8,9,16,17,22,23,20,21,22,23,24,25,16,17,22,23,24,25]
-; AVX512F-FAST-NEXT: vpshufb %ymm13, %ymm5, %ymm5
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,6,7,4,5,6,7,8,9,0,1,6,7,8,9,16,17,22,23,20,21,22,23,24,25,16,17,22,23,24,25]
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm5, %ymm5
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm5, %zmm25
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm15[0],xmm14[1],xmm15[2,3]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm27, %xmm31
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm14, %xmm17
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm6[0],xmm14[1],xmm6[2,3]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm27, %xmm22
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm14, %xmm31
; AVX512F-FAST-NEXT: vpshufb %xmm0, %xmm2, %xmm0
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm6[0,1],ymm8[2],ymm6[3],ymm8[4],ymm6[5,6],ymm8[7],ymm6[8,9],ymm8[10],ymm6[11],ymm8[12],ymm6[13,14],ymm8[15]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm8, %ymm20
-; AVX512F-FAST-NEXT: vpermd %ymm2, %ymm3, %ymm2
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm8[0,1],ymm13[2],ymm8[3],ymm13[4],ymm8[5,6],ymm13[7],ymm8[8,9],ymm13[10],ymm8[11],ymm13[12],ymm8[13,14],ymm13[15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm13, %ymm20
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm8, %ymm30
+; AVX512F-FAST-NEXT: vpermd %ymm2, %ymm18, %ymm2
; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm2, %ymm1
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3,4,5,6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX512F-FAST-NEXT: vpermd %ymm29, %ymm4, %ymm1
-; AVX512F-FAST-NEXT: vpshufb %ymm13, %ymm1, %ymm1
+; AVX512F-FAST-NEXT: vpermd %ymm17, %ymm4, %ymm1
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm1
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm24
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm7[0,1],xmm9[2],xmm7[3]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm10[0],ymm12[1,2],ymm10[3],ymm12[4],ymm10[5],ymm12[6,7],ymm10[8],ymm12[9,10],ymm10[11],ymm12[12],ymm10[13],ymm12[14,15]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm15[0,1],xmm7[2],xmm15[3]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm9[0],ymm10[1,2],ymm9[3],ymm10[4],ymm9[5],ymm10[6,7],ymm9[8],ymm10[9,10],ymm9[11],ymm10[12],ymm9[13],ymm10[14,15]
; AVX512F-FAST-NEXT: vpbroadcastq {{.*#+}} xmm2 = [8,9,2,3,12,13,0,0,8,9,2,3,12,13,0,0]
; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm3
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm27 = <1,4,6,3,6,u,u,u>
@@ -7473,18 +7492,20 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm26 = [0,2,1,3,0,2,5,7]
; AVX512F-FAST-NEXT: vpermd %ymm23, %ymm26, %ymm3
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[0,1,6,7,4,5,6,7,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = [0,1,6,7,4,5,6,7,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31]
+; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm3, %ymm3
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm19
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm3, %zmm1
; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0],ymm15[1,2],ymm4[3],ymm15[4],ymm4[5],ymm15[6,7],ymm4[8],ymm15[9,10],ymm4[11],ymm15[12],ymm4[13],ymm15[14,15]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm23
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0],ymm14[1,2],ymm4[3],ymm14[4],ymm4[5],ymm14[6,7],ymm4[8],ymm14[9,10],ymm4[11],ymm14[12],ymm4[13],ymm14[14,15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm18
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2],xmm3[3]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <0,3,u,u,5,0,2,7>
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, %ymm12
-; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm10[0],ymm12[1],ymm10[2],ymm12[3],ymm10[4,5],ymm12[6],ymm10[7,8],ymm12[9],ymm10[10],ymm12[11],ymm10[12,13],ymm12[14],ymm10[15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, %ymm15
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm29, %ymm10
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm10[0],ymm15[1],ymm10[2],ymm15[3],ymm10[4,5],ymm15[6],ymm10[7,8],ymm15[9],ymm10[10],ymm15[11],ymm10[12,13],ymm15[14],ymm10[15]
; AVX512F-FAST-NEXT: vpermd %ymm5, %ymm4, %ymm5
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [2,3,4,5,10,11,0,1,14,15,14,15,2,3,4,5,18,19,20,21,26,27,16,17,30,31,30,31,18,19,20,21]
; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm5, %ymm5
@@ -7493,9 +7514,9 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4,5,6,7]
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm5 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
; AVX512F-FAST-NEXT: vpternlogq $184, %zmm3, %zmm5, %zmm25
-; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX512F-FAST-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: # ymm3 = ymm14[0],mem[1],ymm14[2,3],mem[4],ymm14[5],mem[6],ymm14[7,8],mem[9],ymm14[10,11],mem[12],ymm14[13],mem[14],ymm14[15]
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3],mem[4],ymm3[5],mem[6],ymm3[7,8],mem[9],ymm3[10,11],mem[12],ymm3[13],mem[14],ymm3[15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm3, %xmm8
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm8 = xmm3[0],xmm8[1,2,3],xmm3[4,5],xmm8[6,7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,0,1,10,11,4,5,14,15,8,9,2,3,12,13]
@@ -7505,34 +7526,35 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm9[0],ymm8[1,2,3,4,5,6,7],ymm9[8],ymm8[9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm25, %zmm21
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, %ymm9
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm11
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm9[0],ymm11[1,2],ymm9[3],ymm11[4],ymm9[5],ymm11[6,7],ymm9[8],ymm11[9,10],ymm9[11],ymm11[12],ymm9[13],ymm11[14,15]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm11[0],ymm12[1,2],ymm11[3],ymm12[4],ymm11[5],ymm12[6,7],ymm11[8],ymm12[9,10],ymm11[11],ymm12[12],ymm11[13],ymm12[14,15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm12, %ymm23
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm11, %ymm25
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm8, %xmm9
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm9[0],xmm8[1],xmm9[2],xmm8[3]
; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm8, %xmm7
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm28, %ymm13
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm8
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm13[0],ymm8[1],ymm13[2],ymm8[3],ymm13[4,5],ymm8[6],ymm13[7,8],ymm8[9],ymm13[10],ymm8[11],ymm13[12,13],ymm8[14],ymm13[15]
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm16, %ymm12
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm13[0],ymm12[1],ymm13[2],ymm12[3],ymm13[4,5],ymm12[6],ymm13[7,8],ymm12[9],ymm13[10],ymm12[11],ymm13[12,13],ymm12[14],ymm13[15]
; AVX512F-FAST-NEXT: vpermd %ymm8, %ymm4, %ymm4
; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm4, %ymm4
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3,4,5,6,7]
; AVX512F-FAST-NEXT: vpternlogq $184, %zmm4, %zmm5, %zmm24
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm31, %xmm4
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm17, %xmm5
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm22, %xmm4
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm31, %xmm5
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3]
; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm4, %xmm2
; AVX512F-FAST-NEXT: vmovdqa64 %ymm20, %ymm4
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm16, %ymm5
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm5
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm5[1,2],ymm4[3],ymm5[4],ymm4[5],ymm5[6,7],ymm4[8],ymm5[9,10],ymm4[11],ymm5[12],ymm4[13],ymm5[14,15]
; AVX512F-FAST-NEXT: vpermd %ymm4, %ymm27, %ymm4
; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm4, %ymm0
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm0[3,4,5,6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-FAST-NEXT: vpermd %ymm29, %ymm26, %ymm2
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[0,1,6,7,4,5,6,7,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31]
+; AVX512F-FAST-NEXT: vpermd %ymm17, %ymm26, %ymm2
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, %ymm4
+; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm2
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm0
-; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX512F-FAST-NEXT: vmovdqu (%rsp), %ymm9 # 32-byte Reload
; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm11[0],ymm9[1],ymm11[2,3],ymm9[4],ymm11[5],ymm9[6],ymm11[7,8],ymm9[9],ymm11[10,11],ymm9[12],ymm11[13],ymm9[14],ymm11[15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm2, %xmm4
@@ -7543,13 +7565,13 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3,4,5,6,7],ymm3[8],ymm2[9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm24, %zmm2
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm10[0,1],ymm12[2],ymm10[3],ymm12[4],ymm10[5,6],ymm12[7],ymm10[8,9],ymm12[10],ymm10[11],ymm12[12],ymm10[13,14],ymm12[15]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm10[0,1],ymm15[2],ymm10[3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8,9],ymm15[10],ymm10[11],ymm15[12],ymm10[13,14],ymm15[15]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <1,3,u,u,6,0,3,5>
; AVX512F-FAST-NEXT: vpermd %ymm3, %ymm4, %ymm3
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = [0,1,6,7,8,9,14,15,0,1,6,7,0,1,6,7,16,17,22,23,24,25,30,31,16,17,22,23,16,17,22,23]
; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm3, %ymm3
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm23, %ymm6
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm15[0],ymm6[1],ymm15[2,3],ymm6[4],ymm15[5],ymm6[6],ymm15[7,8],ymm6[9],ymm15[10,11],ymm6[12],ymm15[13],ymm6[14],ymm15[15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm6
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm14[0],ymm6[1],ymm14[2,3],ymm6[4],ymm14[5],ymm6[6],ymm14[7,8],ymm6[9],ymm14[10,11],ymm6[12],ymm14[13],ymm6[14],ymm14[15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm6, %xmm7
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3,4],xmm7[5,6,7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
@@ -7559,8 +7581,9 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: kmovw %eax, %k1
; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm1 {%k1}
; AVX512F-FAST-NEXT: vextracti64x4 $1, %zmm1, %ymm3
-; AVX512F-FAST-NEXT: vpblendw $181, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm6 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: # ymm6 = mem[0],ymm14[1],mem[2],ymm14[3],mem[4,5],ymm14[6],mem[7,8],ymm14[9],mem[10],ymm14[11],mem[12,13],ymm14[14],mem[15]
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpblendw $181, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: # ymm6 = mem[0],ymm6[1],mem[2],ymm6[3],mem[4,5],ymm6[6],mem[7,8],ymm6[9],mem[10],ymm6[11],mem[12,13],ymm6[14],mem[15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm6, %xmm8
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3],xmm6[4,5,6],xmm8[7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm8 = [6,7,2,3,12,13,6,7,0,1,10,11,4,5,14,15]
@@ -7569,14 +7592,13 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm3[0],ymm6[1,2,3,4,5,6,7],ymm3[8],ymm6[9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm6[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm1, %zmm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm3
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, %ymm6
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm23, %ymm3
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm25, %ymm6
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm6[1],ymm3[2,3],ymm6[4],ymm3[5],ymm6[6],ymm3[7,8],ymm6[9],ymm3[10,11],ymm6[12],ymm3[13],ymm6[14],ymm3[15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm3, %xmm6
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm6[0,1,2],xmm3[3,4],xmm6[5,6,7]
; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm3, %xmm3
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm6
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm13[0,1],ymm6[2],ymm13[3],ymm6[4],ymm13[5,6],ymm6[7],ymm13[8,9],ymm6[10],ymm13[11],ymm6[12],ymm13[13,14],ymm6[15]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm13[0,1],ymm12[2],ymm13[3],ymm12[4],ymm13[5,6],ymm12[7],ymm13[8,9],ymm12[10],ymm13[11],ymm12[12],ymm13[13,14],ymm12[15]
; AVX512F-FAST-NEXT: vpermd %ymm6, %ymm4, %ymm4
; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm4, %ymm4
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
@@ -7598,15 +7620,14 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vmovaps %zmm3, 64(%rdx)
; AVX512F-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-FAST-NEXT: vmovaps %zmm3, (%rdx)
-; AVX512F-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
-; AVX512F-FAST-NEXT: vmovaps %zmm3, 64(%rcx)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm28, 64(%rcx)
; AVX512F-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-FAST-NEXT: vmovaps %zmm3, (%rcx)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm2, 64(%r8)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm21, (%r8)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm0, 64(%r9)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm1, (%r9)
-; AVX512F-FAST-NEXT: addq $712, %rsp # imm = 0x2C8
+; AVX512F-FAST-NEXT: addq $552, %rsp # imm = 0x228
; AVX512F-FAST-NEXT: vzeroupper
; AVX512F-FAST-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
index 893bff29b21e45..eb44ca1800986e 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
@@ -2573,42 +2573,44 @@ define void @load_i16_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqa 208(%rdi), %xmm12
; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %xmm13
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm14 = xmm13[0,1,2,3,4,5],xmm12[6],xmm13[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm14 = xmm14[u,u,u,u,u,u,u,u,u,u,0,1,14,15,12,13]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm14, %xmm14
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm14[5,6,7],ymm4[8,9,10,11,12],ymm14[13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm2[0,1,2],ymm3[3],ymm2[4,5],ymm3[6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm15 = ymm14[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm14 = ymm15[0],ymm14[1,2,3,4,5,6],ymm15[7,8],ymm14[9,10,11,12,13,14],ymm15[15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm15 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm15, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm7, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm11 = xmm15[10,11,6,7,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm11[0],xmm7[0],xmm11[1],xmm7[1],xmm11[2],xmm7[2],xmm11[3],xmm7[3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm11 = ymm14[u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm11[0],ymm4[1,2,3,4,5,6,7],ymm11[8],ymm4[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1],xmm11[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm7 = ymm14[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0],ymm14[1,2,3,4,5,6],ymm7[7,8],ymm14[9,10,11,12,13,14],ymm7[15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm7, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm14, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm15, %xmm11
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm14 = xmm14[10,11,6,7,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm14[0],xmm11[0],xmm14[1],xmm11[1],xmm14[2],xmm11[2],xmm14[3],xmm11[3]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm7[0],ymm4[1,2,3,4,5,6,7],ymm7[8],ymm4[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm7 = xmm11[0,1],xmm7[2,3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm11 = ymm7[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3],xmm5[4],xmm4[5],xmm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm12[0],xmm13[1],xmm12[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,10,11,8,9,6,7,4,5,u,u,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,u,u,u,u,u,u,2,3,0,1,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm12[0],xmm13[1],xmm12[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,4,5,2,3,0,1,14,15,28,29,28,29,28,29,28,29,20,21,18,19,16,17,30,31>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm5, %xmm5
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3,4,5,6,7,8],ymm3[9],ymm2[10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm6, %ymm2, %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5],ymm0[6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,u,u,10,11,6,7,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm1 = ymm2[u,u,u,u,u,u,u,u,4,5,2,3,0,1,14,15,28,29,28,29,28,29,28,29,20,21,18,19,16,17,30,31]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0],ymm4[1,2,3,4,5,6,7],ymm1[8],ymm4[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm4[1,2,3,4,5,6,7],ymm2[8],ymm4[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%rsi)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -5686,16 +5688,16 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX512F-ONLY-SLOW-LABEL: load_i16_stride7_vf32:
; AVX512F-ONLY-SLOW: # %bb.0:
-; AVX512F-ONLY-SLOW-NEXT: pushq %rax
; AVX512F-ONLY-SLOW-NEXT: vmovdqa (%rdi), %ymm2
; AVX512F-ONLY-SLOW-NEXT: vmovdqa 32(%rdi), %ymm3
; AVX512F-ONLY-SLOW-NEXT: vmovdqa 64(%rdi), %ymm12
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 96(%rdi), %ymm4
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0],ymm12[1],ymm4[2,3,4],ymm12[5],ymm4[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 96(%rdi), %ymm11
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0],ymm12[1],ymm11[2,3,4],ymm12[5],ymm11[6,7]
; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[6,7,12,13,2,3,16,17,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1],ymm3[2],ymm2[3,4,5],ymm3[6],ymm2[7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm2, %ymm13
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm3, %ymm4
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm2, %ymm8
; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4],xmm1[5],xmm2[6],xmm1[7]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[0,1,14,15,12,13,10,11,8,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[u,u,u,u,u,u,u,u,u,u,u,u]
@@ -5711,662 +5713,655 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm2
; AVX512F-ONLY-SLOW-NEXT: vmovdqa 192(%rdi), %ymm0
; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,1,0,2]
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} ymm5 = ymm1[0,1,2,1,4,5,6,5]
-; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm5[7]
-; AVX512F-ONLY-SLOW-NEXT: vpbroadcastw 252(%rdi), %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 224(%rdi), %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm7[0,1,0,3]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm7, %xmm14
-; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,7,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpunpckhdq {{.*#+}} xmm5 = xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm1[0,1,2,1,4,5,6,5]
+; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX512F-ONLY-SLOW-NEXT: vpbroadcastw 252(%rdi), %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 224(%rdi), %xmm6
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm6[0,1,0,3]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm6, %xmm13
+; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm5[2],xmm3[2],xmm5[3],xmm3[3]
; AVX512F-ONLY-SLOW-NEXT: movw $992, %ax # imm = 0x3E0
; AVX512F-ONLY-SLOW-NEXT: kmovw %eax, %k1
-; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm5, %zmm2, %zmm21 {%k1}
+; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm3, %zmm2, %zmm21 {%k1}
; AVX512F-ONLY-SLOW-NEXT: vmovdqa 256(%rdi), %ymm5
; AVX512F-ONLY-SLOW-NEXT: vmovdqa 288(%rdi), %ymm6
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm5, %ymm27
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2],xmm2[3],xmm7[4],xmm2[5],xmm7[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 240(%rdi), %xmm15
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 240(%rdi), %xmm14
; AVX512F-ONLY-SLOW-NEXT: vmovdqa 80(%rdi), %xmm2
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm4[0,1],ymm12[2],ymm4[3,4],ymm12[5],ymm4[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm8[0,1,2],ymm2[3],ymm8[4,5,6,7,8,9,10],ymm2[11],ymm8[12,13,14,15]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1],ymm12[2],ymm11[3,4],ymm12[5],ymm11[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6,7,8,9,10],ymm2[11],ymm3[12,13,14,15]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[8,9,6,7,4,5,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm13[0,1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm8, %xmm11
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm8 = xmm11[0],xmm8[1],xmm11[2,3,4,5],xmm8[6],xmm11[7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[2,3,0,1,14,15,12,13,10,11],zero,zero,zero,zero,zero,zero,zero,zero,ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %ymm2, %ymm8, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1,2],ymm4[3],ymm8[4,5],ymm4[6],ymm8[7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm7[0],xmm3[1],xmm7[2,3,4,5],xmm3[6],xmm7[7]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[2,3,0,1,14,15,12,13,10,11],zero,zero,zero,zero,zero,zero,zero,zero,ymm3[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %ymm2, %ymm3, %ymm2
; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1],ymm10[2],ymm9[3,4,5],ymm10[6],ymm9[7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm8
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm8[4],xmm2[5],xmm8[6],xmm2[7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4],xmm2[5],xmm3[6],xmm2[7]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[0,1,0,1,14,15,12,13,10,11,8,9,u,u,u,u]
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,1,1,3,4,5,5,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,5,6,8,9,10,11,12,13,13,14]
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm15[0],xmm14[1],xmm15[2,3,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm14[0],xmm13[1],xmm14[2,3,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm1, %zmm22
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1],ymm12[2,3],ymm4[4,5],ymm12[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm12[2,3],ymm11[4,5],ymm12[6,7]
; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4],ymm1[5,6,7,8,9,10,11],ymm2[12],ymm1[13,14,15]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[10,11,8,9,6,7,20,21],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0],ymm13[1],ymm3[2,3],ymm13[4],ymm3[5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm8
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm8[1],xmm2[2,3,4,5],xmm8[6],xmm2[7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0],ymm8[1],ymm4[2,3],ymm8[4],ymm4[5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2,3,4,5],xmm3[6],xmm2[7]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[4,5,2,3,0,1,14,15,12,13],zero,zero,zero,zero,zero,zero,zero,zero,ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %ymm1, %ymm2, %ymm1
-; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-ONLY-SLOW-NEXT: vporq %ymm1, %ymm2, %ymm19
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1,2],ymm10[3],ymm9[4,5],ymm10[6],ymm9[7]
; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2,3,4,5],xmm1[6],xmm2[7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,1,2,3,0,1,14,15,12,13,10,11,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = <0,1,2,3,0,1,14,15,12,13,10,11,8,9,128,128,128,128,128,128,128,128,128,128,u,u,u,u,u,u,u,u>
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm2, %xmm1, %xmm1
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm2, %ymm28
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[0,1,1,2]
; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[0,1,0,3,4,5,4,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
-; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm14[0],xmm15[0],xmm14[1],xmm15[1],xmm14[2],xmm15[2],xmm14[3],xmm15[3]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm14, %xmm17
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm5[2,1,2,3]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm5, %xmm29
+; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm13[0],xmm14[0],xmm13[1],xmm14[1],xmm13[2],xmm14[2],xmm13[3],xmm14[3]
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm3[2,1,2,3]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm3, %xmm27
; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm1, %zmm24
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm12[3],ymm4[4,5],ymm12[6],ymm4[7]
+; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm1, %zmm23
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1,2],ymm12[3],ymm11[4,5],ymm12[6],ymm11[7]
; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,3,2,3]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[4,5,10,11,0,1,22,23],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0],ymm13[1],ymm3[2,3,4],ymm13[5],ymm3[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm13, %ymm25
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm3, %ymm26
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm8
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm8[0],xmm2[1],xmm8[2],xmm2[3],xmm8[4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0],ymm8[1],ymm4[2,3,4],ymm8[5],ymm4[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm8, %ymm25
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm4, %ymm26
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3],xmm3[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[6,7,4,5,2,3,0,1,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512F-ONLY-SLOW-NEXT: vporq %ymm1, %ymm2, %ymm20
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm23 = ymm0[0,1,1,3]
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm24 = ymm0[0,1,1,3]
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm9[1],ymm10[2,3,4],ymm9[5],ymm10[6,7]
; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2],xmm0[3],xmm1[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,6,7,4,5,2,3,0,1,14,15,u,u,u,u]
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm23[0,1,2,1,4,5,6,5]
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm24[0,1,2,1,4,5,6,5]
; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastw 232(%rdi), %xmm1
-; AVX512F-ONLY-SLOW-NEXT: vpsrlq $48, %xmm15, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpsrlq $48, %xmm14, %xmm2
; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm1, %zmm0, %zmm19
+; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm1, %zmm0, %zmm18
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1],ymm9[2],ymm10[3,4],ymm9[5],ymm10[6,7]
; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3],xmm0[4],xmm1[5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[8,9,8,9,6,7,4,5,2,3,u,u,u,u,u,u]
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 208(%rdi), %xmm13
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 208(%rdi), %xmm2
; AVX512F-ONLY-SLOW-NEXT: vmovdqa 192(%rdi), %xmm1
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,5],xmm13[6],xmm1[7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,6]
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7],ymm0[8,9,10,11,12],ymm2[13,14,15]
-; AVX512F-ONLY-SLOW-NEXT: vpsrld $16, %xmm14, %xmm2
-; AVX512F-ONLY-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm15, %xmm16
-; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm18
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,5],xmm2[6],xmm1[7]
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,7,6]
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7],ymm0[8,9,10,11,12],ymm3[13,14,15]
+; AVX512F-ONLY-SLOW-NEXT: vpsrld $16, %xmm13, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm13, %xmm31
+; AVX512F-ONLY-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm3[4],xmm14[4],xmm3[5],xmm14[5],xmm3[6],xmm14[6],xmm3[7],xmm14[7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm14, %xmm16
+; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm3, %zmm0, %zmm17
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm0[3],xmm3[4],xmm0[5],xmm3[6,7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqa 352(%rdi), %ymm15
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 320(%rdi), %ymm3
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm15[3],ymm3[4,5],ymm15[6],ymm3[7]
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm8 = ymm2[2,3,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm8[6],ymm2[7,8,9,10,11,12,13],ymm8[14],ymm2[15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,2,2,1]
-; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[0,1,14,15,12,13,10,11,4,5,6,7,8,9,2,3,16,17,30,31,28,29,26,27,20,21,22,23,24,25,18,19]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm7 = xmm2[0,1,2],xmm7[3,4,5,6],xmm2[7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 384(%rdi), %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 416(%rdi), %ymm8
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1],ymm7[2],ymm8[3,4],ymm7[5],ymm8[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm11
-; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,1,3,1,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 320(%rdi), %ymm4
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm4[0,1,2],ymm15[3],ymm4[4,5],ymm15[6],ymm4[7]
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm13 = ymm8[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm13[6],ymm8[7,8,9,10,11,12,13],ymm13[14],ymm8[15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,2,2,1]
+; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[0,1,14,15,12,13,10,11,4,5,6,7,8,9,2,3,16,17,30,31,28,29,26,27,20,21,22,23,24,25,18,19]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm8[0,1,2],xmm3[3,4,5,6],xmm8[7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 384(%rdi), %ymm8
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 416(%rdi), %ymm13
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1],ymm8[2],ymm13[3,4],ymm8[5],ymm13[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm7
+; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,3,1,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
+; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm0, %ymm30
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm27, %ymm5
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm0, %ymm29
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2],ymm5[3],ymm6[4,5],ymm5[6],ymm6[7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1,2,3],xmm0[4],xmm2[5],xmm0[6],xmm2[7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0],ymm3[1],ymm15[2,3],ymm3[4],ymm15[5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm11 = ymm2[2,3,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm11[0],ymm2[1,2,3,4,5,6],ymm11[7,8],ymm2[9,10,11,12,13,14],ymm11[15]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0,1,2,3],xmm0[4],xmm3[5],xmm0[6],xmm3[7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0],ymm4[1],ymm15[2,3],ymm4[4],ymm15[5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm7[0],ymm3[1,2,3,4,5,6],ymm7[7,8],ymm3[9,10,11,12,13,14],ymm7[15]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,12,13,10,11,8,9,6,7,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[2,3,0,1,14,15,12,13,4,5,4,5,4,5,4,5,18,19,16,17,30,31,28,29,20,21,20,21,20,21,20,21]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3,4,5,6],xmm2[7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1],ymm7[2,3],ymm8[4,5],ymm7[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm11
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[2,1,2,3]
-; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,1,0,2,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
-; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1],xmm2[2],xmm11[2],xmm2[3],xmm11[3]
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm0, %ymm28
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0],ymm3[1],ymm15[2,3,4],ymm3[5],ymm15[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[2,3,0,1,14,15,12,13,4,5,4,5,4,5,4,5,18,19,16,17,30,31,28,29,20,21,20,21,20,21,20,21]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0,1,2],xmm0[3,4,5,6],xmm3[7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1],ymm8[2,3],ymm13[4,5],ymm8[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[2,1,2,3]
+; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,0,2,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
+; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1],xmm3[2],xmm7[2],xmm3[3],xmm7[3]
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm0, %ymm30
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0],ymm4[1],ymm15[2,3,4],ymm4[5],ymm15[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm0[2,3,0,1]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,u,u,6,7,20,21,u,u,16,17,30,31,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4,5,6,7,8],ymm2[9],ymm0[10,11,12,13,14,15]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm6[2],ymm5[3,4,5],ymm6[6],ymm5[7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm11
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm11[4],xmm2[5],xmm11[6],xmm2[7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[0,1,2,3,0,1,14,15,12,13,10,11,8,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm7[3],ymm8[4,5],ymm7[6],ymm8[7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm11
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4,5,6,7,8],ymm3[9],ymm0[10,11,12,13,14,15]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm6[2],ymm5[3,4,5],ymm6[6],ymm5[7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm7[4],xmm3[5],xmm7[6],xmm3[7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm28, %ymm7
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %ymm7, %ymm3, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2],ymm8[3],ymm13[4,5],ymm8[6],ymm13[7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm7
; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,4]
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[0,1,2,1]
-; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5,5,7]
-; AVX512F-ONLY-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm11[4],xmm0[5],xmm11[5],xmm0[6],xmm11[6],xmm0[7],xmm11[7]
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,1,2,1]
+; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,5,7]
+; AVX512F-ONLY-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm0, %ymm31
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm0, %ymm28
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm9[1],ymm10[2,3],ymm9[4],ymm10[5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3,4,5],xmm2[6],xmm0[7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1],xmm0[2,3,4,5],xmm3[6],xmm0[7]
; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,0,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,6,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm23[0,1,2,0,4,5,6,4]
-; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,7,4,8,9,10,11,12,13,15,12]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm29, %xmm2
-; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
-; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm27
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm24[0,1,2,0,4,5,6,4]
+; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,7,4,8,9,10,11,12,13,15,12]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm27, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
+; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm3, %zmm0, %zmm27
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm6[3],ymm5[4,5],ymm6[6],ymm5[7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3,4,5],xmm0[6],xmm2[7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1],ymm8[2],ymm7[3,4,5],ymm8[6],ymm7[7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm14
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm14[4],xmm2[5],xmm14[6],xmm2[7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0],xmm0[1],xmm3[2,3,4,5],xmm0[6],xmm3[7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1],ymm13[2],ymm8[3,4,5],ymm13[6],ymm8[7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm14
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm14[4],xmm3[5],xmm14[6],xmm3[7]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,2,3,0,1,14,15,12,13,10,11],zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1],ymm3[2],ymm15[3,4],ymm3[5],ymm15[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1],ymm4[2],ymm15[3,4],ymm4[5],ymm15[6,7]
; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm14 = ymm14[1,1,2,0]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm14 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm14[0,1,22,23,28,29,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX512F-ONLY-SLOW-NEXT: vpor %ymm0, %ymm14, %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm14[0,1,2],ymm2[3,4,5,6,7],ymm14[8,9,10],ymm2[11,12,13,14,15]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm0, %ymm29
+; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm14[0,1,2],ymm3[3,4,5,6,7],ymm14[8,9,10],ymm3[11,12,13,14,15]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm0, %ymm24
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1],ymm9[2,3],ymm10[4,5],ymm9[6,7]
; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm9
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm9[0,1,2],xmm0[3],xmm9[4],xmm0[5],xmm9[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm13[0],xmm1[1],xmm13[2,3,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2,3,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[8,9,10,11,8,9,6,7,4,5,u,u,u,u,u,u]
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm17, %xmm1
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm31, %xmm1
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm16, %xmm2
; AVX512F-ONLY-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm1, %zmm0, %zmm23
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1],ymm4[2],ymm12[3,4,5],ymm4[6],ymm12[7]
+; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm1, %zmm0, %zmm16
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1],ymm11[2],ymm12[3,4,5],ymm11[6],ymm12[7]
; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,0,1,14,15,u,u,10,11,24,25,24,25,24,25,24,25,16,17,30,31,u,u,26,27]
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7,8,9,10,11,12,13],ymm1[14],ymm0[15]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm25, %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm26, %ymm11
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm2[2],ymm11[3,4],ymm2[5],ymm11[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm10
-; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[3,1,2,3,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm25, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm26, %ymm7
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm3[2],ymm7[3,4],ymm3[5],ymm7[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1],xmm1[2],xmm10[2],xmm1[3],xmm10[3]
+; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2],ymm8[3],ymm7[4,5],ymm8[6],ymm7[7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm10
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm10[0],xmm0[1],xmm10[2,3,4,5],xmm0[6],xmm10[7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm15[0,1],ymm3[2,3],ymm15[4,5],ymm3[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm13 = ymm15[0,1,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm10[0,1,2],ymm13[3],ymm10[4,5,6,7,8,9,10],ymm13[11],ymm10[12,13,14,15]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm6[0],ymm5[1],ymm6[2,3],ymm5[4],ymm6[5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm13, %xmm14
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0],xmm14[1],xmm13[2,3,4,5],xmm14[6],xmm13[7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm13[3],ymm8[4,5],ymm13[6],ymm8[7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3,4,5],xmm0[6],xmm2[7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0,1],ymm4[2,3],ymm15[4,5],ymm4[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm15[0,1,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2],ymm10[3],ymm2[4,5,6,7,8,9,10],ymm10[11],ymm2[12,13,14,15]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm6[0],ymm5[1],ymm6[2,3],ymm5[4],ymm6[5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm10, %xmm14
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0],xmm14[1],xmm10[2,3,4,5],xmm14[6],xmm10[7]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,2,3,0,1,14,15,12,13,10,11]
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[10,11,24,25,22,23,20,21],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5,6,7],ymm10[8,9,10],ymm0[11,12,13,14,15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5,2,3,0,1,14,15,12,13],zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpor %ymm10, %ymm13, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm10[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0],ymm7[1],ymm8[2,3],ymm7[4],ymm8[5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm10
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm10[1],xmm0[2,3,4,5],xmm10[6],xmm0[7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm15[0,1,2],ymm3[3],ymm15[4,5],ymm3[6],ymm15[7]
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm14 = ymm10[2,3,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm14[4],ymm10[5,6,7,8,9,10,11],ymm14[12],ymm10[13,14,15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[10,11,24,25,22,23,20,21],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7],ymm2[8,9,10],ymm0[11,12,13,14,15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[0,1,2,3,4,5,2,3,0,1,14,15,12,13],zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpor %ymm2, %ymm10, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0],ymm8[1],ymm13[2,3],ymm8[4],ymm13[5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3,4,5],xmm2[6],xmm0[7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0,1,2],ymm4[3],ymm15[4,5],ymm4[6],ymm15[7]
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm14 = ymm2[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4],ymm2[5,6,7,8,9,10,11],ymm14[12],ymm2[13,14,15]
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm6[0],ymm5[1],ymm6[2,3,4],ymm5[5],ymm6[6,7]
; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm14, %xmm9
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0],xmm14[1],xmm9[2],xmm14[3],xmm9[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,4,5,2,3,0,1,14,15,12,13]
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[12,13,26,27,24,25,22,23],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5,6,7],ymm10[8,9,10],ymm0[11,12,13,14,15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[12,13,26,27,24,25,22,23],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7],ymm2[8,9,10],ymm0[11,12,13,14,15]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm9 = xmm9[0,1,2,3,6,7,4,5,2,3,0,1,14,15],zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpor %ymm10, %ymm9, %ymm9
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm9 = ymm12[0,1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7]
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm9[2,3,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm9 = ymm10[0],ymm9[1,2,3,4,5,6],ymm10[7,8],ymm9[9,10,11,12,13,14],ymm10[15]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm11[0,1],ymm2[2,3],ymm11[4,5],ymm2[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm10, %xmm14
+; AVX512F-ONLY-SLOW-NEXT: vpor %ymm2, %ymm9, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0,1,2],ymm11[3],ymm12[4,5],ymm11[6],ymm12[7]
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm2[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm9[0],ymm2[1,2,3,4,5,6],ymm9[7,8],ymm2[9,10,11,12,13,14],ymm9[15]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm9 = ymm7[0,1],ymm3[2,3],ymm7[4,5],ymm3[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm9, %xmm14
; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm14 = xmm14[2,1,2,3]
; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm14 = xmm14[0,2,2,3,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[2,1,2,3]
-; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[1,3,2,3,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm10[0],xmm14[0],xmm10[1],xmm14[1],xmm10[2],xmm14[2],xmm10[3],xmm14[3]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1],ymm9[2,3,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0],ymm12[1],ymm4[2,3],ymm12[4],ymm4[5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm4[2,3,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm9[1],ymm4[2,3,4,5,6,7,8],ymm9[9],ymm4[10,11,12,13,14,15]
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm30, %zmm0, %zmm9
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm28, %zmm0, %zmm12
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm31, %zmm0, %zmm14
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm29, %zmm0, %zmm25
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm13, %zmm0, %zmm13
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[2,1,2,3]
+; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm9[0],xmm14[0],xmm9[1],xmm14[1],xmm9[2],xmm14[2],xmm9[3],xmm14[3]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1],ymm2[2,3,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm9 = ymm11[0],ymm12[1],ymm11[2,3],ymm12[4],ymm11[5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm11 = ymm9[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm9 = ymm9[0],ymm11[1],ymm9[2,3,4,5,6,7,8],ymm11[9],ymm9[10,11,12,13,14,15]
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm29, %zmm0, %zmm11
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm30, %zmm0, %zmm12
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm28, %zmm0, %zmm14
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm24, %zmm0, %zmm25
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm10, %zmm0, %zmm10
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm26
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1,2],ymm2[3],ymm11[4,5],ymm2[6],ymm11[7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,6,4,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
-; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,7,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm4[u,u,u,u,u,u,u,u,4,5,2,3,0,1,14,15,28,29,28,29,28,29,28,29,20,21,18,19,16,17,30,31]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm0[0,1],ymm2[2,3,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0],ymm7[1],ymm8[2,3,4],ymm7[5],ymm8[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm4
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm4[0],xmm0[1],xmm4[2],xmm0[3],xmm4[4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1],ymm5[2],ymm6[3,4],ymm5[5],ymm6[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,3,1,3,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
-; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,0,2,3,4,5,6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm15[2],ymm3[3,4,5],ymm15[6],ymm3[7]
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,3,1]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[4,5,10,11,u,u,u,u,u,u,u,u,0,1,14,15,20,21,26,27,16,17,26,27,16,17,20,21,16,17,30,31]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0],ymm4[1,2],ymm3[3,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1,2],ymm3[3],ymm7[4,5],ymm3[6],ymm7[7]
+; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,6,4,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
+; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,7,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = ymm9[u,u,u,u,u,u,u,u,4,5,2,3,0,1,14,15,28,29,28,29,28,29,28,29,20,21,18,19,16,17,30,31]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0,1],ymm3[2,3,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0],ymm8[1],ymm13[2,3,4],ymm8[5],ymm13[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm8
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm8[0],xmm0[1],xmm8[2],xmm0[3],xmm8[4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1],ymm5[2],ymm6[3,4],ymm5[5],ymm6[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,3,1,3,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
+; AVX512F-ONLY-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,0,2,3,4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm15[2],ymm4[3,4,5],ymm15[6],ymm4[7]
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,3,1]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[4,5,10,11,u,u,u,u,u,u,u,u,0,1,14,15,20,21,26,27,16,17,26,27,16,17,20,21,16,17,30,31]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0],ymm5[1,2],ymm4[3,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,6,7,4,5,2,3,0,1,14,15]
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7],ymm3[8,9,10],ymm0[11,12,13,14,15]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7],ymm4[8,9,10],ymm0[11,12,13,14,15]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm21, %zmm3, %zmm9
-; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm22
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm22, %zmm3, %zmm12
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm24 # 64-byte Folded Reload
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm4 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm21, %zmm4, %zmm11
+; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm5, %zmm22
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm22, %zmm4, %zmm12
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm4 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm19, %zmm4, %zmm23
; AVX512F-ONLY-SLOW-NEXT: movw $-512, %ax # imm = 0xFE00
; AVX512F-ONLY-SLOW-NEXT: kmovw %eax, %k1
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm14, %zmm24 {%k1}
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm20, %zmm3, %zmm27
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm14, %zmm23 {%k1}
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm20, %zmm4, %zmm27
; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm25, %zmm27 {%k1}
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %zmm19, %zmm3, %zmm1
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm13, %zmm1 {%k1}
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm9, (%rsi)
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %zmm18, %zmm4, %zmm1
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm10, %zmm1 {%k1}
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm11, (%rsi)
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm12, (%rdx)
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm24, (%rcx)
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm23, (%rcx)
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm27, (%r8)
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm1, (%r9)
; AVX512F-ONLY-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %zmm18, %zmm3, %zmm10
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm26, %zmm10 {%k1}
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm10, (%rax)
-; AVX512F-ONLY-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %zmm23, %zmm3, %zmm2
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm0, %zmm2 {%k1}
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %zmm17, %zmm4, %zmm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm26, %zmm2 {%k1}
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm2, (%rax)
-; AVX512F-ONLY-SLOW-NEXT: popq %rax
+; AVX512F-ONLY-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %zmm16, %zmm4, %zmm3
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm0, %zmm3 {%k1}
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm3, (%rax)
; AVX512F-ONLY-SLOW-NEXT: vzeroupper
; AVX512F-ONLY-SLOW-NEXT: retq
;
; AVX512F-ONLY-FAST-LABEL: load_i16_stride7_vf32:
; AVX512F-ONLY-FAST: # %bb.0:
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 320(%rdi), %zmm28
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 64(%rdi), %zmm1
-; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [2,6,9,13,2,6,9,13]
-; AVX512F-ONLY-FAST-NEXT: # ymm2 = mem[0,1,0,1]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 128(%rdi), %zmm23
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 320(%rdi), %zmm30
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 64(%rdi), %zmm31
+; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [2,6,9,13,2,6,9,13]
+; AVX512F-ONLY-FAST-NEXT: # ymm1 = mem[0,1,0,1]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 128(%rdi), %zmm25
; AVX512F-ONLY-FAST-NEXT: vbroadcasti32x4 {{.*#+}} ymm16 = [2,5,9,12,2,5,9,12]
; AVX512F-ONLY-FAST-NEXT: # ymm16 = mem[0,1,2,3,0,1,2,3]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} ymm19 = [10,3,6,15,12,13,6,15]
-; AVX512F-ONLY-FAST-NEXT: vbroadcasti32x4 {{.*#+}} ymm27 = [3,6,10,13,3,6,10,13]
-; AVX512F-ONLY-FAST-NEXT: # ymm27 = mem[0,1,2,3,0,1,2,3]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 384(%rdi), %zmm18
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} ymm26 = <1,u,u,u,5,8,12,15>
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} ymm18 = [10,3,6,15,12,13,6,15]
+; AVX512F-ONLY-FAST-NEXT: vbroadcasti32x4 {{.*#+}} ymm19 = [3,6,10,13,3,6,10,13]
+; AVX512F-ONLY-FAST-NEXT: # ymm19 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 384(%rdi), %zmm14
; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <2,6,9,u,13,u,u,u>
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm1, %zmm0, %zmm14
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm18, %zmm2, %zmm13
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm31, %zmm0, %zmm10
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm14, %zmm1, %zmm12
; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <1,u,u,u,4,8,11,15>
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm28, %zmm0, %zmm2
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm30, %zmm0, %zmm15
; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <2,5,9,u,12,u,u,u>
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm1, %zmm0, %zmm10
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm18, %zmm16, %zmm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <0,u,u,u,4,7,11,14>
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm28, %zmm3, %zmm6
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [8,1,12,5,12,5,14,15]
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm1, %zmm3, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm23, %zmm27, %zmm4
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 192(%rdi), %ymm25
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm25[0,1,0,2]
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastd {{.*#+}} ymm7 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm7, %ymm8, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm7, %ymm30
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm31, %zmm0, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm14, %zmm16, %zmm7
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <0,u,u,u,4,7,11,14>
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm30, %zmm0, %zmm2
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [8,1,12,5,12,5,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm31, %zmm0, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm25, %zmm19, %zmm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 192(%rdi), %ymm28
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm11 = ymm28[0,1,0,2]
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastd {{.*#+}} ymm1 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm1, %ymm11, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm1, %ymm23
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[0,1,2,3,0,1,6,7,8,9,14,15,12,13,14,15,16,17,18,19,16,17,22,23,24,25,30,31,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm4[0,1,2,3,4,5,6],ymm5[7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[6,7,12,13,2,3,16,17,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rdi), %ymm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rdi), %ymm4
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm3[0,1],ymm4[2],ymm3[3,4,5],ymm4[6],ymm3[7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm11, %xmm12
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2,3],xmm12[4],xmm11[5],xmm12[6],xmm11[7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm11[0,1,14,15,12,13,10,11,8,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm11[u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vporq %ymm7, %ymm11, %ymm22
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastw 252(%rdi), %xmm7
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 224(%rdi), %xmm12
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm12[u,u,u,u,u,u,u,u,0,1,14,15,12,13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vpunpckhdq {{.*#+}} xmm7 = xmm11[2],xmm7[2],xmm11[3],xmm7[3]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm4[0,1,2,3,4,5,6],ymm5[7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[6,7,12,13,2,3,16,17,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rdi), %ymm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rdi), %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm4[0,1],ymm5[2],ymm4[3,4,5],ymm5[6],ymm4[7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm8, %xmm9
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1,2,3],xmm9[4],xmm8[5],xmm9[6],xmm8[7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[0,1,14,15,12,13,10,11,8,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm8[u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vporq %ymm0, %ymm8, %ymm22
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastw 252(%rdi), %xmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 224(%rdi), %xmm13
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm13[u,u,u,u,u,u,u,u,0,1,14,15,12,13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm8[2],xmm0[2],xmm8[3],xmm0[3]
; AVX512F-ONLY-FAST-NEXT: movw $992, %ax # imm = 0x3E0
; AVX512F-ONLY-FAST-NEXT: kmovw %eax, %k1
-; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm7, %zmm5, %zmm22 {%k1}
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 256(%rdi), %ymm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 288(%rdi), %ymm7
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm7[0,1],ymm5[2,3],ymm7[4,5],ymm5[6,7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm11, %xmm15
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm15[0,1,2],xmm11[3],xmm15[4],xmm11[5],xmm15[6,7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[u,u,u,u,u,u,10,11,8,9,6,7,4,5,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[0,1,6,7,8,9,14,15,8,9,14,15,4,5,2,3,16,17,22,23,24,25,30,31,24,25,30,31,20,21,18,19]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm6[0,1,2],xmm11[3,4,5,6],xmm6[7]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm6[4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastq {{.*#+}} ymm6 = [16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm6, %ymm0, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1,2,3,4,5],ymm0[6,7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm29
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[0,1,6,7,8,9,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm3[0,1,2],ymm4[3],ymm3[4,5],ymm4[6],ymm3[7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm10, %xmm11
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm11[0],xmm10[1],xmm11[2,3,4,5],xmm10[6],xmm11[7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm10 = ymm10[2,3,0,1,14,15,12,13,10,11],zero,zero,zero,zero,zero,zero,zero,zero,ymm10[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %ymm0, %ymm10, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 160(%rdi), %ymm11
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 128(%rdi), %ymm10
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1],ymm11[2],ymm10[3,4,5],ymm11[6],ymm10[7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm15
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm15[4],xmm0[5],xmm15[6],xmm0[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 240(%rdi), %xmm15
+; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm0, %zmm6, %zmm22 {%k1}
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 256(%rdi), %ymm6
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 288(%rdi), %ymm8
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1],ymm6[2,3],ymm8[4,5],ymm6[6,7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm9
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm9[0,1,2],xmm0[3],xmm9[4],xmm0[5],xmm9[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,10,11,8,9,6,7,4,5,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[0,1,6,7,8,9,14,15,8,9,14,15,4,5,2,3,16,17,22,23,24,25,30,31,24,25,30,31,20,21,18,19]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3,4,5,6],xmm2[7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastq {{.*#+}} ymm2 = [16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm7, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm24
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm5[3],ymm4[4,5],ymm5[6],ymm4[7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm7
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm7[0],xmm0[1],xmm7[2,3,4,5],xmm0[6],xmm7[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <2,3,0,1,14,15,12,13,10,11,128,128,128,128,128,128,128,128,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm1, %ymm0, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[0,1,6,7,8,9,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vporq %ymm3, %ymm0, %ymm20
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 160(%rdi), %ymm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 128(%rdi), %ymm9
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1],ymm3[2],ymm9[3,4,5],ymm3[6],ymm9[7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm7
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm7[4],xmm0[5],xmm7[6],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 240(%rdi), %xmm7
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,14,15,12,13,10,11,8,9,u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,22,23,28,29]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm8[7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm8 = xmm15[0],xmm12[1],xmm15[2,3,4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[2,3,0,1,14,15,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm0, %zmm24
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2],ymm5[3],ymm7[4,5],ymm5[6],ymm7[7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm8
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm8[0,1,2,3],xmm0[4],xmm8[5],xmm0[6],xmm8[7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,22,23,28,29]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm11[7]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm7[0],xmm13[1],xmm7[2,3,4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm1, %xmm11, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm27
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm6[3],ymm8[4,5],ymm6[6],ymm8[7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4],xmm1[5],xmm0[6],xmm1[7]
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,12,13,10,11,8,9,6,7,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[2,3,4,5,10,11,12,13,0,1,0,1,0,1,0,1,18,19,20,21,26,27,28,29,16,17,16,17,16,17,16,17]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3,4,5,6],xmm2[7]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27,28,29]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm20
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm14[2,3,4,5,10,11,16,17],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm2, %xmm8
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm8[1],xmm2[2,3,4,5],xmm8[6],xmm2[7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[4,5,2,3,0,1,14,15,12,13],zero,zero,zero,zero,zero,zero,zero,zero,ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vporq %ymm0, %ymm2, %ymm31
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm11[3],ymm10[4,5],ymm11[6],ymm10[7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm2
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3,4,5],xmm0[6],xmm2[7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,0,1,14,15,12,13,10,11,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm15[2,3,4,5,10,11,12,13,0,1,0,1,0,1,0,1,18,19,20,21,26,27,28,29,16,17,16,17,16,17,16,17]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3,4,5,6],xmm1[7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm12[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27,28,29]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm26
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[2,3,4,5,10,11,16,17],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm1, %xmm10
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3,4,5],xmm10[6],xmm1[7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[4,5,2,3,0,1,14,15,12,13],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vporq %ymm0, %ymm1, %ymm21
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3,4,5],xmm0[6],xmm1[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <0,1,2,3,0,1,14,15,12,13,10,11,8,9,128,128,128,128,128,128,128,128,128,128,u,u,u,u,u,u,u,u>
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastq {{.*#+}} ymm2 = [2,5,2,5,2,5,2,5]
-; AVX512F-ONLY-FAST-NEXT: vpermd %ymm25, %ymm2, %ymm2
-; AVX512F-ONLY-FAST-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm12[0],xmm15[0],xmm12[1],xmm15[1],xmm12[2],xmm15[2],xmm12[3],xmm15[3]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm2[8,9,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm0, %zmm14
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1],ymm7[2],ymm5[3,4,5],ymm7[6],ymm5[7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm8
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm8[4],xmm0[5],xmm8[6],xmm0[7]
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm28, %zmm26, %zmm8
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,0,1,14,15,12,13,10,11,8,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm8 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm8[2,3,16,17,22,23,24,25,30,31,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %ymm0, %ymm8, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm18, %zmm27, %zmm8
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm6, %ymm8, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm6[6,7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm27
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0],ymm3[1],ymm4[2,3,4],ymm3[5],ymm4[6,7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm6
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm6[0],xmm0[1],xmm6[2],xmm0[3],xmm6[4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm1, %zmm19, %zmm6
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[4,5,10,11,0,1,22,23],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[6,7,4,5,2,3,0,1,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vporq %ymm6, %ymm0, %ymm21
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0],ymm10[1],ymm11[2,3],ymm10[4],ymm11[5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm6
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm6[1],xmm0[2,3,4,5],xmm6[6],xmm0[7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,2,3,0,1,14,15,12,13,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm25[0,1,1,3]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,24,25]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[12,13,10,11,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm6, %zmm19
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 416(%rdi), %ymm2
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 384(%rdi), %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1],ymm2[2],ymm8[3,4,5],ymm2[6],ymm8[7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm6, %xmm9
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3],xmm9[4],xmm6[5],xmm9[6],xmm6[7]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm5[0,1,2],ymm7[3],ymm5[4,5],ymm7[6],ymm5[7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm9, %xmm13
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm9 = xmm13[0],xmm9[1],xmm13[2,3,4,5],xmm9[6],xmm13[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = [2,11,2,11,12,5,8,9]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u,u,u,u,u,0,1,14,15,12,13,10,11,8,9]
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm28, %zmm13, %zmm13
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm13 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm13[0,1,22,23,28,29,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm13[0,1,2],ymm6[3,4,5,6,7],ymm13[8,9,10],ymm6[11,12,13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm9[0,1,2,3,2,3,0,1,14,15,12,13,10,11],zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpor %ymm13, %ymm9, %ymm9
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm6, %ymm17
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm11[0],ymm10[1],ymm11[2,3,4],ymm10[5],ymm11[6,7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm9, %xmm10
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0],xmm9[1],xmm10[2],xmm9[3],xmm10[4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm9[0,1,6,7,4,5,2,3,0,1,14,15,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm30, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm6, %ymm0, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5,6],ymm0[7]
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastw 232(%rdi), %xmm9
-; AVX512F-ONLY-FAST-NEXT: vpsrlq $48, %xmm15, %xmm10
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3]
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm23, %zmm16, %zmm10
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = <0,3,7,10,14,u,u,u>
-; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm9, %zmm0, %zmm9
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = <0,3,3,u,0,3,7,u>
-; AVX512F-ONLY-FAST-NEXT: vpermd %ymm25, %ymm13, %ymm13
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,u,u,u,u,u,u,u,0,1,6,7,8,9,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm10 = ymm10[0,1,0,1,6,7,8,9,14,15,u,u,u,u,u,u,16,17,16,17,22,23,24,25,30,31,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm13[5,6,7],ymm10[8,9,10,11,12],ymm13[13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vpsrld $16, %xmm12, %xmm13
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm13 = xmm13[4],xmm15[4],xmm13[5],xmm15[5],xmm13[6],xmm15[6],xmm13[7],xmm15[7]
-; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm13, %zmm10, %zmm26
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm13 = [8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm13, %xmm0, %xmm10
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX512F-ONLY-FAST-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm1, %zmm11, %zmm10
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31>
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm6, %ymm10, %ymm10
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm6, %ymm18
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm10[2,3,4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm30 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm0, %zmm30, %zmm9
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm2[3],ymm8[4,5],ymm2[6],ymm8[7]
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastq {{.*#+}} ymm10 = [2,5,2,5,2,5,2,5]
+; AVX512F-ONLY-FAST-NEXT: vpermd %ymm28, %ymm10, %ymm10
+; AVX512F-ONLY-FAST-NEXT: vpshufhw {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm10[7]
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm13[0],xmm7[0],xmm13[1],xmm7[1],xmm13[2],xmm7[2],xmm13[3],xmm7[3]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm10[8,9,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm0, %zmm15
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1],ymm8[2],ymm6[3,4,5],ymm8[6],ymm6[7]
; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm11
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm11[0],xmm0[1],xmm11[2,3,4,5],xmm0[6],xmm11[7]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm7[0],ymm5[1],ymm7[2,3],ymm5[4],ymm7[5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm11, %xmm6
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm11[0],xmm6[1],xmm11[2,3,4,5],xmm6[6],xmm11[7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,2,3,0,1,14,15,12,13,10,11]
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = <2,u,u,u,6,9,13,u>
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm28, %zmm11, %zmm11
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm11 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm11[2,3,16,17,22,23,24,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm11[0,1,2],ymm0[3,4,5,6,7],ymm11[8,9,10],ymm0[11,12,13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,2,3,0,1,14,15,12,13],zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpor %ymm6, %ymm11, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm4[0,1],ymm3[2,3],ymm4[4,5],ymm3[6,7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm11, %xmm0
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm13, %xmm0, %xmm10
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm29, %zmm0, %zmm16
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm20, %zmm0, %zmm13
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm27, %zmm0, %zmm27
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm17, %zmm0, %zmm29
-; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [2,6,9,13,2,6,9,13]
-; AVX512F-ONLY-FAST-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm23, %zmm0, %zmm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} ymm23 = <0,4,7,11,14,u,u,u>
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm0, %zmm17
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm1, %zmm23, %zmm6
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[10,11,6,7,4,5,6,7,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,u,u,u,u,2,3,4,5,10,11,12,13,18,19,18,19,18,19,18,19,18,19,20,21,26,27,28,29]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm10[0,1],ymm6[2,3,4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm2[0],ymm8[1],ymm2[2,3],ymm8[4],ymm2[5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm6, %xmm10
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0],xmm10[1],xmm6[2,3,4,5],xmm10[6],xmm6[7]
-; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm10 = [0,4,7,0,0,4,7,0]
-; AVX512F-ONLY-FAST-NEXT: # ymm10 = mem[0,1,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpermd %ymm25, %ymm10, %ymm10
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm10 = ymm10[u,u,u,u,u,u,u,u,u,u,2,3,4,5,10,11,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[4,5,2,3,4,5,10,11,12,13,u,u,u,u,u,u,20,21,18,19,20,21,26,27,28,29,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm10[5,6,7],ymm0[8,9,10,11,12],ymm10[13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm10 = xmm12[4],xmm15[4],xmm12[5],xmm15[5],xmm12[6],xmm15[6],xmm12[7],xmm15[7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[8,9,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <3,u,u,u,6,10,13,u>
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm10, %zmm0, %zmm12
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0],ymm5[1],ymm7[2,3,4],ymm5[5],ymm7[6,7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm10
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm10[0],xmm0[1],xmm10[2],xmm0[3],xmm10[4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm28, %zmm15, %zmm10
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u,u,u,u,u,4,5,2,3,0,1,14,15,12,13]
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[0,1,18,19,20,21,26,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm10[0,1,2],ymm6[3,4,5,6,7],ymm10[8,9,10],ymm6[11,12,13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,4,5,2,3,0,1,14,15],zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpor %ymm0, %ymm10, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <1,4,8,11,15,u,u,u>
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm10
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm1, %zmm6, %zmm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm18, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm11[4],xmm0[5],xmm11[6],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm31, %zmm18, %zmm11
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm12 = <1,u,u,u,5,8,12,15>
; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm1, %ymm0, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm3[3],ymm4[4,5],ymm3[6],ymm4[7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm1, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,u,10,11,6,7,4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,4,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm8[1],ymm2[2,3,4],ymm8[5],ymm2[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm30, %zmm12, %zmm1
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[2,3,16,17,22,23,24,25,30,31,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %ymm1, %ymm0, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm14, %zmm19, %zmm1
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm17
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm11[4,5,10,11,0,1,22,23],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0],ymm4[1],ymm5[2,3,4],ymm4[5],ymm5[6,7]
; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3],xmm2[4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [0,1,10,3,14,7,10,3]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm5[2],ymm7[3,4],ymm5[5],ymm7[6,7]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
-; AVX512F-ONLY-FAST-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,3,1,3,4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[8,9,8,9,4,5,6,7,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm28, %zmm2, %zmm2
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[4,5,10,11,u,u,u,u,u,u,u,u,0,1,14,15,20,21,26,27,16,17,26,27,16,17,20,21,16,17,30,31]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm3[1,2],ymm2[3,4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,6,7,4,5,2,3,0,1,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[6,7,4,5,2,3,0,1,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vporq %ymm0, %ymm1, %ymm18
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0],ymm9[1],ymm3[2,3],ymm9[4],ymm3[5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5],xmm1[6],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,2,3,0,1,14,15,12,13,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm11 = ymm28[0,1,1,3]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,24,25]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm10[12,13,10,11,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm19
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 416(%rdi), %ymm14
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 384(%rdi), %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1],ymm14[2],ymm2[3,4,5],ymm14[6],ymm2[7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4],xmm0[5],xmm1[6],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1,2],ymm8[3],ymm6[4,5],ymm8[6],ymm6[7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm1, %xmm10
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm10[0],xmm1[1],xmm10[2,3,4,5],xmm1[6],xmm10[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = [2,11,2,11,12,5,8,9]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,0,1,14,15,12,13,10,11,8,9]
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm30, %zmm10, %zmm10
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[0,1,22,23,28,29,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5,6,7],ymm10[8,9,10],ymm0[11,12,13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,1,2,3,2,3,0,1,14,15,12,13,10,11],zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpor %ymm1, %ymm10, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <0,3,7,10,14,u,u,u>
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0],ymm9[1],ymm3[2,3,4],ymm9[5],ymm3[6,7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm1, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2],xmm1[3],xmm3[4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm31, %zmm0, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,1,6,7,4,5,2,3,0,1,14,15,u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7],ymm2[8,9,10],ymm1[11,12,13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm1
-; AVX512F-ONLY-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm22, %zmm2, %zmm16
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm3, %zmm24
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm24, %zmm2, %zmm13
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm31, %zmm30, %zmm14
-; AVX512F-ONLY-FAST-NEXT: movw $-512, %di # imm = 0xFE00
-; AVX512F-ONLY-FAST-NEXT: kmovw %edi, %k1
-; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm27, %zmm14 {%k1}
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm21, %zmm30, %zmm19
-; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm29, %zmm19 {%k1}
-; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm17, %zmm9 {%k1}
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm26, %zmm30, %zmm11
-; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm10, %zmm11 {%k1}
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm16, (%rsi)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm13, (%rdx)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm14, (%rcx)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm23, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm11, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7]
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastw 232(%rdi), %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpsrlq $48, %xmm7, %xmm9
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm5[0,1],ymm4[2],ymm5[3,4],ymm4[5],ymm5[6,7]
+; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm3, %zmm1, %zmm23
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm12 = [8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm12, %xmm9, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm9, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = <u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31>
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm9, %ymm0, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm25, %zmm16, %zmm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm16 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm0, %zmm16, %zmm23
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <0,3,3,u,0,3,7,u>
+; AVX512F-ONLY-FAST-NEXT: vpermd %ymm28, %ymm0, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,0,1,6,7,8,9,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[0,1,0,1,6,7,8,9,14,15,u,u,u,u,u,u,16,17,16,17,22,23,24,25,30,31,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7],ymm1[8,9,10,11,12],ymm0[13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpsrld $16, %xmm13, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
+; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm1, %zmm0, %zmm29
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm14[3],ymm2[4,5],ymm14[6],ymm2[7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3,4,5],xmm0[6],xmm1[7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0],ymm6[1],ymm8[2,3],ymm6[4],ymm8[5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm1, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3,4,5],xmm3[6],xmm1[7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,2,3,0,1,14,15,12,13,10,11]
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <2,u,u,u,6,9,13,u>
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm30, %zmm3, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[2,3,16,17,22,23,24,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7],ymm3[8,9,10],ymm0[11,12,13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,2,3,0,1,14,15,12,13],zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpor %ymm3, %ymm1, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm12, %xmm1, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm26, %zmm0, %zmm26
+; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [2,6,9,13,2,6,9,13]
+; AVX512F-ONLY-FAST-NEXT: # ymm3 = mem[0,1,0,1]
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm25, %zmm3, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm12 = <0,4,7,11,14,u,u,u>
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm17, %zmm0, %zmm17
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm31, %zmm12, %zmm12
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[10,11,6,7,4,5,6,7,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm12[u,u,u,u,u,u,u,u,2,3,4,5,10,11,12,13,18,19,18,19,18,19,18,19,18,19,20,21,26,27,28,29]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0],ymm2[1],ymm14[2,3],ymm2[4],ymm14[5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5],xmm1[6],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [0,4,7,0,0,4,7,0]
+; AVX512F-ONLY-FAST-NEXT: # ymm1 = mem[0,1,0,1]
+; AVX512F-ONLY-FAST-NEXT: vpermd %ymm28, %ymm1, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,u,u,2,3,4,5,10,11,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[4,5,2,3,4,5,10,11,12,13,u,u,u,u,u,u,20,21,18,19,20,21,26,27,28,29,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm1[5,6,7],ymm3[8,9,10,11,12],ymm1[13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm13[4],xmm7[4],xmm13[5],xmm7[5],xmm13[6],xmm7[6],xmm13[7],xmm7[7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[8,9,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm1, %zmm7
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0],ymm6[1],ymm8[2,3,4],ymm6[5],ymm8[6,7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm1, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2],xmm1[3],xmm3[4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <3,u,u,u,6,10,13,u>
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm10, %zmm0, %zmm10
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm30, %zmm3, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,4,5,2,3,0,1,14,15,12,13]
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[0,1,18,19,20,21,26,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7],ymm3[8,9,10],ymm0[11,12,13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,1,2,3,6,7,4,5,2,3,0,1,14,15],zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpor %ymm3, %ymm1, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <1,4,8,11,15,u,u,u>
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm0, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm31, %zmm1, %zmm1
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm9, %ymm1, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5],ymm4[6],ymm5[7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,u,u,u,u,10,11,6,7,4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,4,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3,4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0],ymm2[1],ymm14[2,3,4],ymm2[5],ymm14[6,7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm2, %xmm4
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0],xmm2[1],xmm4[2],xmm2[3],xmm4[4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1],ymm6[2],ymm8[3,4],ymm6[5],ymm8[6,7]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX512F-ONLY-FAST-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,3,1,3,4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[8,9,8,9,4,5,6,7,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = [0,1,10,3,14,7,10,3]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm30, %zmm5, %zmm5
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[4,5,10,11,u,u,u,u,u,u,u,u,0,1,14,15,20,21,26,27,16,17,26,27,16,17,20,21,16,17,30,31]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0],ymm4[1,2],ymm5[3,4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,6,7,4,5,2,3,0,1,14,15]
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3,4,5,6,7],ymm4[8,9,10],ymm2[11,12,13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm2
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm4 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm22, %zmm4, %zmm24
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm20, %zmm27
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm27, %zmm4, %zmm26
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm21, %zmm16, %zmm15
+; AVX512F-ONLY-FAST-NEXT: movw $-512, %ax # imm = 0xFE00
+; AVX512F-ONLY-FAST-NEXT: kmovw %eax, %k1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm17, %zmm15 {%k1}
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm18, %zmm16, %zmm19
+; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm10, %zmm19 {%k1}
+; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm3, %zmm23 {%k1}
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm29, %zmm16, %zmm12
+; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm0, %zmm12 {%k1}
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm24, (%rsi)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm26, (%rdx)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm15, (%rcx)
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm19, (%r8)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm9, (%r9)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm11, (%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm23, (%r9)
; AVX512F-ONLY-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm12, %zmm30, %zmm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm1, %zmm0 {%k1}
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm0, (%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm12, (%rax)
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm7, %zmm16, %zmm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm2, %zmm1 {%k1}
+; AVX512F-ONLY-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm1, (%rax)
; AVX512F-ONLY-FAST-NEXT: vzeroupper
; AVX512F-ONLY-FAST-NEXT: retq
;
; AVX512DQ-SLOW-LABEL: load_i16_stride7_vf32:
; AVX512DQ-SLOW: # %bb.0:
-; AVX512DQ-SLOW-NEXT: pushq %rax
; AVX512DQ-SLOW-NEXT: vmovdqa (%rdi), %ymm2
; AVX512DQ-SLOW-NEXT: vmovdqa 32(%rdi), %ymm3
; AVX512DQ-SLOW-NEXT: vmovdqa 64(%rdi), %ymm1
-; AVX512DQ-SLOW-NEXT: vmovdqa 96(%rdi), %ymm4
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0],ymm1[1],ymm4[2,3,4],ymm1[5],ymm4[6,7]
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm4, %ymm7
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm1, %ymm12
+; AVX512DQ-SLOW-NEXT: vmovdqa 96(%rdi), %ymm10
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm1[1],ymm10[2,3,4],ymm1[5],ymm10[6,7]
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm1, %ymm4
; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[6,7,12,13,2,3,16,17,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1],ymm3[2],ymm2[3,4,5],ymm3[6],ymm2[7]
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm3, %ymm10
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm3, %ymm12
; AVX512DQ-SLOW-NEXT: vmovdqa %ymm2, %ymm11
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4],xmm1[5],xmm2[6],xmm1[7]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[0,1,14,15,12,13,10,11,8,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %ymm0, %ymm1, %ymm6
+; AVX512DQ-SLOW-NEXT: vporq %ymm0, %ymm1, %ymm18
; AVX512DQ-SLOW-NEXT: vmovdqa 128(%rdi), %ymm8
; AVX512DQ-SLOW-NEXT: vmovdqa 160(%rdi), %ymm9
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2],ymm8[3],ymm9[4,5],ymm8[6],ymm9[7]
@@ -6382,29 +6377,24 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX512DQ-SLOW-NEXT: vpbroadcastw 252(%rdi), %xmm3
-; AVX512DQ-SLOW-NEXT: vmovdqa 224(%rdi), %xmm5
-; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm5[0,1,0,3]
-; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
-; AVX512DQ-SLOW-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX512DQ-SLOW-NEXT: vmovdqa 224(%rdi), %xmm6
+; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm6[0,1,0,3]
+; AVX512DQ-SLOW-NEXT: vmovdqa %xmm6, %xmm13
+; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
+; AVX512DQ-SLOW-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm5[2],xmm3[2],xmm5[3],xmm3[3]
; AVX512DQ-SLOW-NEXT: movw $992, %ax # imm = 0x3E0
; AVX512DQ-SLOW-NEXT: kmovw %eax, %k1
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm3, %zmm2, %zmm6 {%k1}
-; AVX512DQ-SLOW-NEXT: vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-SLOW-NEXT: vmovdqa 256(%rdi), %ymm3
-; AVX512DQ-SLOW-NEXT: vmovdqa 288(%rdi), %ymm4
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1],ymm3[2,3],ymm4[4,5],ymm3[6,7]
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm4, %ymm6
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm3, %ymm25
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm14 = xmm3[0,1,2],xmm2[3],xmm3[4],xmm2[5],xmm3[6,7]
-; AVX512DQ-SLOW-NEXT: vmovdqa 240(%rdi), %xmm15
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm3, %zmm2, %zmm18 {%k1}
+; AVX512DQ-SLOW-NEXT: vmovdqa 256(%rdi), %ymm6
+; AVX512DQ-SLOW-NEXT: vmovdqa 288(%rdi), %ymm5
+; AVX512DQ-SLOW-NEXT: vmovdqa 240(%rdi), %xmm14
; AVX512DQ-SLOW-NEXT: vmovdqa 80(%rdi), %xmm2
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm12[2],ymm7[3,4],ymm12[5],ymm7[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0,1],ymm4[2],ymm10[3,4],ymm4[5],ymm10[6,7]
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6,7,8,9,10],ymm2[11],ymm3[12,13,14,15]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[8,9,6,7,4,5,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1,2],ymm10[3],ymm11[4,5],ymm10[6],ymm11[7]
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3,4,5],xmm3[6],xmm4[7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1,2],ymm12[3],ymm11[4,5],ymm12[6],ymm11[7]
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm7[0],xmm3[1],xmm7[2,3,4,5],xmm3[6],xmm7[7]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[2,3,0,1,14,15,12,13,10,11],zero,zero,zero,zero,zero,zero,zero,zero,ymm3[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-SLOW-NEXT: vpor %ymm2, %ymm3, %ymm2
; AVX512DQ-SLOW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
@@ -6416,87 +6406,92 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,1,1,3,4,5,5,7]
; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,5,6,8,9,10,11,12,13,13,14]
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm15[0],xmm5[1],xmm15[2,3,4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm14[0],xmm13[1],xmm14[2,3,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm1, %zmm20
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm12[2,3],ymm7[4,5],ymm12[6,7]
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm1, %zmm21
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1],ymm4[2,3],ymm10[4,5],ymm4[6,7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4],ymm1[5,6,7,8,9,10,11],ymm2[12],ymm1[13,14,15]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[10,11,8,9,6,7,20,21],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0],ymm11[1],ymm10[2,3],ymm11[4],ymm10[5,6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0],ymm11[1],ymm12[2,3],ymm11[4],ymm12[5,6,7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2,3,4,5],xmm3[6],xmm2[7]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[4,5,2,3,0,1,14,15,12,13],zero,zero,zero,zero,zero,zero,zero,zero,ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vporq %ymm1, %ymm2, %ymm18
+; AVX512DQ-SLOW-NEXT: vporq %ymm1, %ymm2, %ymm19
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2],ymm9[3],ymm8[4,5],ymm9[6],ymm8[7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2,3,4,5],xmm1[6],xmm2[7]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,1,2,3,0,1,14,15,12,13,10,11,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = <0,1,2,3,0,1,14,15,12,13,10,11,8,9,128,128,128,128,128,128,128,128,128,128,u,u,u,u,u,u,u,u>
+; AVX512DQ-SLOW-NEXT: vpshufb %xmm2, %xmm1, %xmm1
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm2, %ymm25
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[0,1,1,2]
; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[0,1,0,3,4,5,4,7]
; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
-; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm15[0],xmm5[1],xmm15[1],xmm5[2],xmm15[2],xmm5[3],xmm15[3]
+; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm13[0],xmm14[0],xmm13[1],xmm14[1],xmm13[2],xmm14[2],xmm13[3],xmm14[3]
; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm3[2,1,2,3]
-; AVX512DQ-SLOW-NEXT: vmovdqa %xmm3, %xmm13
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm3, %xmm26
; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm1, %zmm24
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1,2],ymm12[3],ymm7[4,5],ymm12[6],ymm7[7]
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm1, %zmm22
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm4[3],ymm10[4,5],ymm4[6],ymm10[7]
; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,3,2,3]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[4,5,10,11,0,1,22,23],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0],ymm11[1],ymm10[2,3,4],ymm11[5],ymm10[6,7]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm11, %ymm26
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0],ymm11[1],ymm12[2,3,4],ymm11[5],ymm12[6,7]
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm12, %ymm15
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3],xmm3[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[6,7,4,5,2,3,0,1,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vporq %ymm1, %ymm2, %ymm19
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm22 = ymm0[0,1,1,3]
+; AVX512DQ-SLOW-NEXT: vporq %ymm1, %ymm2, %ymm17
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm23 = ymm0[0,1,1,3]
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0],ymm8[1],ymm9[2,3,4],ymm8[5],ymm9[6,7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2],xmm0[3],xmm1[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,6,7,4,5,2,3,0,1,14,15,u,u,u,u]
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm22[0,1,2,1,4,5,6,5]
+; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm23[0,1,2,1,4,5,6,5]
; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
; AVX512DQ-SLOW-NEXT: vpbroadcastw 232(%rdi), %xmm1
-; AVX512DQ-SLOW-NEXT: vpsrlq $48, %xmm15, %xmm2
+; AVX512DQ-SLOW-NEXT: vpsrlq $48, %xmm14, %xmm2
; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm1, %zmm0, %zmm21
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm1, %zmm0, %zmm20
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1],ymm8[2],ymm9[3,4],ymm8[5],ymm9[6,7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3],xmm0[4],xmm1[5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[8,9,8,9,6,7,4,5,2,3,u,u,u,u,u,u]
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512DQ-SLOW-NEXT: vmovdqa 208(%rdi), %xmm2
-; AVX512DQ-SLOW-NEXT: vmovdqa 192(%rdi), %xmm3
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3,4,5],xmm2[6],xmm3[7]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm3, %xmm17
+; AVX512DQ-SLOW-NEXT: vmovdqa 192(%rdi), %xmm7
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm7[0,1,2,3,4,5],xmm2[6],xmm7[7]
; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm2, %xmm29
; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,6]
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
-; AVX512DQ-SLOW-NEXT: vpsrld $16, %xmm5, %xmm1
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm5, %xmm27
-; AVX512DQ-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm15[4],xmm1[5],xmm15[5],xmm1[6],xmm15[6],xmm1[7],xmm15[7]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm15, %xmm28
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm1, %zmm0, %zmm23
-; AVX512DQ-SLOW-NEXT: vmovdqa 352(%rdi), %ymm11
-; AVX512DQ-SLOW-NEXT: vmovdqa 320(%rdi), %ymm15
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2],ymm11[3],ymm15[4,5],ymm11[6],ymm15[7]
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6],ymm0[7,8,9,10,11,12,13],ymm2[14],ymm0[15]
-; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm14[0,2,2,1]
-; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,14,15,12,13,10,11,4,5,6,7,8,9,2,3,16,17,30,31,28,29,26,27,20,21,22,23,24,25,18,19]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm0[0,1,2],xmm2[3,4,5,6],xmm0[7]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vmovdqa 384(%rdi), %ymm14
-; AVX512DQ-SLOW-NEXT: vmovdqa 416(%rdi), %ymm5
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm14[2],ymm5[3,4],ymm14[5],ymm5[6,7]
+; AVX512DQ-SLOW-NEXT: vpsrld $16, %xmm13, %xmm1
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm13, %xmm27
+; AVX512DQ-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm14[4],xmm1[5],xmm14[5],xmm1[6],xmm14[6],xmm1[7],xmm14[7]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm14, %xmm28
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm1, %zmm0, %zmm24
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm5, %ymm12
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3],ymm5[4,5],ymm6[6,7]
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3],xmm1[4],xmm0[5],xmm1[6,7]
+; AVX512DQ-SLOW-NEXT: vmovdqa 352(%rdi), %ymm1
+; AVX512DQ-SLOW-NEXT: vmovdqa 320(%rdi), %ymm14
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0,1,2],ymm1[3],ymm14[4,5],ymm1[6],ymm14[7]
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6],ymm2[7,8,9,10,11,12,13],ymm3[14],ymm2[15]
+; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,1]
+; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[0,1,14,15,12,13,10,11,4,5,6,7,8,9,2,3,16,17,30,31,28,29,26,27,20,21,22,23,24,25,18,19]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3,4,5,6],xmm2[7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vmovdqa 384(%rdi), %ymm5
+; AVX512DQ-SLOW-NEXT: vmovdqa 416(%rdi), %ymm13
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1],ymm5[2],ymm13[3,4],ymm5[5],ymm13[6,7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
@@ -6505,18 +6500,17 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm0, %ymm30
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm25, %ymm4
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2],ymm4[3],ymm6[4,5],ymm4[6],ymm6[7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1,2],ymm6[3],ymm12[4,5],ymm6[6],ymm12[7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1,2,3],xmm0[4],xmm2[5],xmm0[6],xmm2[7]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0],ymm15[1],ymm11[2,3],ymm15[4],ymm11[5,6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0],ymm14[1],ymm1[2,3],ymm14[4],ymm1[5,6,7]
; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3,4,5,6],ymm3[7,8],ymm2[9,10,11,12,13,14],ymm3[15]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,12,13,10,11,8,9,6,7,u,u]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[2,3,0,1,14,15,12,13,4,5,4,5,4,5,4,5,18,19,16,17,30,31,28,29,20,21,20,21,20,21,20,21]
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3,4,5,6],xmm2[7]
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm14[2,3],ymm5[4,5],ymm14[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1],ymm5[2,3],ymm13[4,5],ymm5[6,7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,2,4,5,6,7]
@@ -6525,17 +6519,18 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm0, %ymm16
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0],ymm15[1],ymm11[2,3,4],ymm15[5],ymm11[6,7]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm0, %ymm31
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,u,u,6,7,20,21,u,u,16,17,30,31,u,u,u,u,u,u,u,u]
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4,5,6,7,8],ymm2[9],ymm0[10,11,12,13,14,15]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1],ymm6[2],ymm4[3,4,5],ymm6[6],ymm4[7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1],ymm12[2],ymm6[3,4,5],ymm12[6],ymm6[7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4],xmm2[5],xmm3[6],xmm2[7]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[0,1,2,3,0,1,14,15,12,13,10,11,8,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm25, %ymm3
+; AVX512DQ-SLOW-NEXT: vpshufb %ymm3, %ymm2, %ymm2
; AVX512DQ-SLOW-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm2
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm14[3],ymm5[4,5],ymm14[6],ymm5[7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2],ymm5[3],ymm13[4,5],ymm5[6],ymm13[7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,4]
; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
@@ -6544,31 +6539,31 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm25 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm18, %zmm25, %zmm24
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm19, %zmm25, %zmm22
; AVX512DQ-SLOW-NEXT: movw $-512, %ax # imm = 0xFE00
; AVX512DQ-SLOW-NEXT: kmovw %eax, %k1
-; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm24 {%k1}
+; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm22 {%k1}
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0],ymm8[1],ymm9[2,3],ymm8[4],ymm9[5,6,7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3,4,5],xmm2[6],xmm0[7]
; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,0,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,6,7]
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm22[0,1,2,0,4,5,6,4]
+; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm23[0,1,2,0,4,5,6,4]
; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,7,4,8,9,10,11,12,13,15,12]
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
-; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm13[0,1,2,3,6,5,6,7]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm26, %xmm2
+; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm18
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm6[3],ymm4[4,5],ymm6[6],ymm4[7]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm6, %ymm31
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm19
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2],ymm12[3],ymm6[4,5],ymm12[6],ymm6[7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3,4,5],xmm0[6],xmm2[7]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0,1],ymm5[2],ymm14[3,4,5],ymm5[6],ymm14[7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm13[2],ymm5[3,4,5],ymm13[6],ymm5[7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4],xmm2[5],xmm3[6],xmm2[7]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,2,3,0,1,14,15,12,13,10,11],zero,zero
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1],ymm15[2],ymm11[3,4],ymm15[5],ymm11[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1],ymm14[2],ymm1[3,4],ymm14[5],ymm1[6,7]
; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[1,1,2,0]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[0,1,22,23,28,29,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX512DQ-SLOW-NEXT: vpor %ymm3, %ymm0, %ymm0
@@ -6577,446 +6572,434 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm19, %zmm25, %zmm18
-; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm18 {%k1}
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm12, %ymm1
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm6
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1],ymm7[2],ymm12[3,4,5],ymm7[6],ymm12[7]
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm17, %zmm25, %zmm19
+; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm19 {%k1}
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm10[2],ymm4[3,4,5],ymm10[6],ymm4[7]
; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,0,1,14,15,u,u,10,11,24,25,24,25,24,25,24,25,16,17,30,31,u,u,26,27]
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6],ymm0[7,8,9,10,11,12,13],ymm2[14],ymm0[15]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm26, %ymm13
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1],ymm13[2],ymm10[3,4],ymm13[5],ymm10[6,7]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm10, %ymm22
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm15, %ymm23
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0,1],ymm11[2],ymm15[3,4],ymm11[5],ymm15[6,7]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm11, %ymm26
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,2,2,3,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0,1,2],ymm5[3],ymm14[4,5],ymm5[6],ymm14[7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm11 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2],ymm13[3],ymm5[4,5],ymm13[6],ymm5[7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2,3,4,5],xmm2[6],xmm3[7]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1],ymm15[2,3],ymm11[4,5],ymm15[6,7]
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm12 = ymm11[0,1,0,1]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm12[3],ymm3[4,5,6,7,8,9,10],ymm12[11],ymm3[12,13,14,15]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm31, %ymm7
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm7[0],ymm4[1],ymm7[2,3],ymm4[4],ymm7[5,6,7]
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm12, %xmm10
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm10 = xmm12[0],xmm10[1],xmm12[2,3,4,5],xmm10[6],xmm12[7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1],ymm14[2,3],ymm1[4,5],ymm14[6,7]
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm15 = ymm1[0,1,0,1]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm15[3],ymm3[4,5,6,7,8,9,10],ymm15[11],ymm3[12,13,14,15]
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm12, %ymm0
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm15 = ymm12[0],ymm6[1],ymm12[2,3],ymm6[4],ymm12[5,6,7]
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm15, %xmm12
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm12 = xmm15[0],xmm12[1],xmm15[2,3,4,5],xmm12[6],xmm15[7]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,2,3,0,1,14,15,12,13,10,11]
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[10,11,24,25,22,23,20,21],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[0,1,2,3,4,5,2,3,0,1,14,15,12,13],zero,zero
-; AVX512DQ-SLOW-NEXT: vpor %ymm3, %ymm10, %ymm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm12 = xmm12[0,1,2,3,4,5,2,3,0,1,14,15,12,13],zero,zero
+; AVX512DQ-SLOW-NEXT: vpor %ymm3, %ymm12, %ymm3
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vpternlogq $226, %zmm21, %zmm25, %zmm0
-; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm2, %zmm0, %zmm0 {%k1}
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0],ymm14[1],ymm5[2,3],ymm14[4],ymm5[5,6,7]
+; AVX512DQ-SLOW-NEXT: vpternlogq $226, %zmm20, %zmm25, %zmm11
+; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm2, %zmm0, %zmm11 {%k1}
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0],ymm5[1],ymm13[2,3],ymm5[4],ymm13[5,6,7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2,3,4,5],xmm3[6],xmm2[7]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1,2],ymm15[3],ymm11[4,5],ymm15[6],ymm11[7]
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm3[2,3,0,1]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm10[4],ymm3[5,6,7,8,9,10,11],ymm10[12],ymm3[13,14,15]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm7[0],ymm4[1],ymm7[2,3,4],ymm4[5],ymm7[6,7]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm4, %ymm21
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm10, %xmm12
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm10 = xmm12[0],xmm10[1],xmm12[2],xmm10[3],xmm12[4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2],ymm14[3],ymm1[4,5],ymm14[6],ymm1[7]
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm12 = ymm3[2,3,0,1]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4],ymm3[5,6,7,8,9,10,11],ymm12[12],ymm3[13,14,15]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm0[0],ymm6[1],ymm0[2,3,4],ymm6[5],ymm0[6,7]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm0, %ymm20
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm12, %xmm15
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm12 = xmm15[0],xmm12[1],xmm15[2],xmm12[3],xmm15[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,4,5,2,3,0,1,14,15,12,13]
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[12,13,26,27,24,25,22,23],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[0,1,2,3,6,7,4,5,2,3,0,1,14,15],zero,zero
-; AVX512DQ-SLOW-NEXT: vpor %ymm3, %ymm10, %ymm3
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2],ymm6[3],ymm1[4,5],ymm6[6],ymm1[7]
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm1, %ymm3
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm2[2,3,0,1]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm10[0],ymm2[1,2,3,4,5,6],ymm10[7,8],ymm2[9,10,11,12,13,14],ymm10[15]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm12 = xmm12[0,1,2,3,6,7,4,5,2,3,0,1,14,15],zero,zero
+; AVX512DQ-SLOW-NEXT: vpor %ymm3, %ymm12, %ymm3
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm0, %ymm17
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2],ymm10[3],ymm4[4,5],ymm10[6],ymm4[7]
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm4, %ymm3
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm12 = ymm2[2,3,0,1]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm12[0],ymm2[1,2,3,4,5,6],ymm12[7,8],ymm2[9,10,11,12,13,14],ymm12[15]
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3],ymm9[4,5],ymm8[6,7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm8, %xmm9
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm10 = xmm9[0,1,2],xmm8[3],xmm9[4],xmm8[5],xmm9[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm12 = xmm9[0,1,2],xmm8[3],xmm9[4],xmm8[5],xmm9[6,7]
; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm30, %zmm0, %zmm8
-; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm16, %zmm0, %zmm9
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm29, %xmm1
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm17, %xmm4
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm1[0],xmm4[1],xmm1[2,3,4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[8,9,10,11,8,9,6,7,4,5,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
-; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm10[0,1,2,3,4],ymm4[5,6,7],ymm10[8,9,10,11,12],ymm4[13,14,15]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm27, %xmm1
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm28, %xmm10
-; AVX512DQ-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm10 = xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
-; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[2,1,2,3]
-; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,3,2,3,4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm10, %zmm4, %zmm19
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm22, %ymm4
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm4[0,1],ymm13[2,3],ymm4[4,5],ymm13[6,7]
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm10, %xmm12
+; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm31, %zmm0, %zmm9
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm29, %xmm0
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm7 = xmm0[0],xmm7[1],xmm0[2,3,4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm12 = xmm12[8,9,10,11,8,9,6,7,4,5,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
+; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,1,0,3]
+; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,4,7]
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm12[0,1,2,3,4],ymm7[5,6,7],ymm12[8,9,10,11,12],ymm7[13,14,15]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm27, %xmm0
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm28, %xmm4
+; AVX512DQ-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm12 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm12 = xmm12[2,1,2,3]
+; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[0,3,2,3,4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm12, %zmm7, %zmm27
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm23, %ymm7
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm26, %ymm0
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm7[0,1],ymm0[2,3],ymm7[4,5],ymm0[6,7]
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm12, %xmm15
+; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm15 = xmm15[2,1,2,3]
+; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[0,2,2,3,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm12 = xmm12[2,1,2,3]
-; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[0,2,2,3,4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[2,1,2,3]
-; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[1,3,2,3,4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm10[0],xmm12[0],xmm10[1],xmm12[1],xmm10[2],xmm12[2],xmm10[3],xmm12[3]
+; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[1,3,2,3,4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm12[0],xmm15[0],xmm12[1],xmm15[1],xmm12[2],xmm15[2],xmm12[3],xmm15[3]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1],ymm2[2,3,4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vpternlogq $226, %zmm23, %zmm25, %zmm2
-; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm7, %zmm0, %zmm2 {%k1}
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0],ymm3[1],ymm6[2,3],ymm3[4],ymm6[5,6,7]
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm7[1],ymm3[2,3,4,5,6,7,8],ymm7[9],ymm3[10,11,12,13,14,15]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm13[3],ymm4[4,5],ymm13[6],ymm4[7]
-; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm10 = xmm7[0,1,2,3,6,4,6,7]
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm7
-; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,1,2,1]
-; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,5,7,6,7]
-; AVX512DQ-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm7 = xmm10[4],xmm7[4],xmm10[5],xmm7[5],xmm10[6],xmm7[6],xmm10[7],xmm7[7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0,1],ymm2[2,3,4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vpternlogq $226, %zmm24, %zmm25, %zmm2
+; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm17, %zmm0, %zmm2 {%k1}
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0],ymm3[1],ymm10[2,3],ymm3[4],ymm10[5,6,7]
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4,5,6,7,8],ymm4[9],ymm3[10,11,12,13,14,15]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0,1,2],ymm0[3],ymm7[4,5],ymm0[6],ymm7[7]
+; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm10 = xmm4[0,1,2,3,6,4,6,7]
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm4
+; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
+; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,7,6,7]
+; AVX512DQ-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm10[4],xmm4[4],xmm10[5],xmm4[5],xmm10[6],xmm4[6],xmm10[7],xmm4[7]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[u,u,u,u,u,u,u,u,4,5,2,3,0,1,14,15,28,29,28,29,28,29,28,29,20,21,18,19,16,17,30,31]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm3[2,3,4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm5[0],ymm14[1],ymm5[2,3,4],ymm14[5],ymm5[6,7]
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm10
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm7 = xmm10[0],xmm7[1],xmm10[2],xmm7[3],xmm10[4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm21, %ymm1
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm31, %ymm4
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm4[0,1],ymm1[2],ymm4[3,4],ymm1[5],ymm4[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm13[0],ymm5[1],ymm13[2,3,4],ymm5[5],ymm13[6,7]
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm10
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm10[0],xmm4[1],xmm10[2],xmm4[3],xmm10[4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm20, %ymm0
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm0[0,1],ymm6[2],ymm0[3,4],ymm6[5],ymm0[6,7]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,3,1,3,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
; AVX512DQ-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,0,2,3,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm11[2],ymm15[3,4,5],ymm11[6],ymm15[7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1],ymm1[2],ymm14[3,4,5],ymm1[6],ymm14[7]
; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,3,1]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[4,5,10,11,u,u,u,u,u,u,u,u,0,1,14,15,20,21,26,27,16,17,26,27,16,17,20,21,16,17,30,31]
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],ymm5[1,2],ymm1[3,4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm7[u,u,u,u,u,u,6,7,4,5,2,3,0,1,14,15]
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm1[0,1,2],ymm5[3,4,5,6,7],ymm1[8,9,10],ymm5[11,12,13,14,15]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vpternlogq $226, %zmm19, %zmm25, %zmm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,6,7,4,5,2,3,0,1,14,15]
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm1[0,1,2],ymm4[3,4,5,6,7],ymm1[8,9,10],ymm4[11,12,13,14,15]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vpternlogq $226, %zmm27, %zmm25, %zmm3
; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm1, %zmm0, %zmm3 {%k1}
; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm8 # 64-byte Folded Reload
-; AVX512DQ-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
-; AVX512DQ-SLOW-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm20
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm20, %zmm1, %zmm9
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm18, %zmm1, %zmm8
+; AVX512DQ-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512DQ-SLOW-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm21
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm21, %zmm1, %zmm9
; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm9, (%rdx)
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm24, (%rcx)
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm18, (%r8)
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm0, (%r9)
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm22, (%rcx)
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm19, (%r8)
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm11, (%r9)
; AVX512DQ-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm2, (%rax)
; AVX512DQ-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm3, (%rax)
-; AVX512DQ-SLOW-NEXT: popq %rax
; AVX512DQ-SLOW-NEXT: vzeroupper
; AVX512DQ-SLOW-NEXT: retq
;
; AVX512DQ-FAST-LABEL: load_i16_stride7_vf32:
; AVX512DQ-FAST: # %bb.0:
-; AVX512DQ-FAST-NEXT: vmovdqa64 320(%rdi), %zmm27
-; AVX512DQ-FAST-NEXT: vmovdqa64 64(%rdi), %zmm29
-; AVX512DQ-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [2,6,9,13,2,6,9,13]
-; AVX512DQ-FAST-NEXT: # ymm1 = mem[0,1,0,1]
-; AVX512DQ-FAST-NEXT: vmovdqa64 128(%rdi), %zmm24
-; AVX512DQ-FAST-NEXT: vbroadcasti32x4 {{.*#+}} ymm30 = [2,5,9,12,2,5,9,12]
-; AVX512DQ-FAST-NEXT: # ymm30 = mem[0,1,2,3,0,1,2,3]
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [10,3,6,15,12,13,6,15]
-; AVX512DQ-FAST-NEXT: vpermd %zmm29, %zmm2, %zmm0
-; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vbroadcasti32x4 {{.*#+}} ymm21 = [3,6,10,13,3,6,10,13]
-; AVX512DQ-FAST-NEXT: # ymm21 = mem[0,1,2,3,0,1,2,3]
-; AVX512DQ-FAST-NEXT: vmovdqa64 384(%rdi), %zmm17
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} ymm19 = <1,u,u,u,5,8,12,15>
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <2,6,9,u,13,u,u,u>
-; AVX512DQ-FAST-NEXT: vpermd %zmm29, %zmm2, %zmm12
-; AVX512DQ-FAST-NEXT: vpermd %zmm17, %zmm1, %zmm15
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <1,u,u,u,4,8,11,15>
-; AVX512DQ-FAST-NEXT: vpermd %zmm27, %zmm2, %zmm2
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <2,5,9,u,12,u,u,u>
-; AVX512DQ-FAST-NEXT: vpermd %zmm29, %zmm3, %zmm3
-; AVX512DQ-FAST-NEXT: vpermd %zmm17, %zmm30, %zmm9
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <0,u,u,u,4,7,11,14>
-; AVX512DQ-FAST-NEXT: vpermd %zmm27, %zmm4, %zmm8
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = [8,1,12,5,12,5,14,15]
-; AVX512DQ-FAST-NEXT: vpermd %zmm29, %zmm4, %zmm4
-; AVX512DQ-FAST-NEXT: vpermd %zmm24, %zmm21, %zmm5
-; AVX512DQ-FAST-NEXT: vmovdqa64 192(%rdi), %ymm26
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm13 = ymm26[0,1,0,2]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[0,1,2,3,0,1,6,7,8,9,14,15,12,13,14,15,16,17,18,19,16,17,22,23,24,25,30,31,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm5[0,1,2,3,4,5,6],ymm6[7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm4[6,7,12,13,2,3,16,17,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512DQ-FAST-NEXT: vmovdqa 32(%rdi), %ymm1
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm0[0,1],ymm1[2],ymm0[3,4,5],ymm1[6],ymm0[7]
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm1, %ymm4
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm0, %ymm1
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm10, %xmm11
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2,3],xmm11[4],xmm10[5],xmm11[6],xmm10[7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm10 = ymm10[0,1,14,15,12,13,10,11,8,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vporq %ymm7, %ymm10, %ymm22
-; AVX512DQ-FAST-NEXT: vpbroadcastw 252(%rdi), %xmm7
-; AVX512DQ-FAST-NEXT: vmovdqa 224(%rdi), %xmm0
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm0[u,u,u,u,u,u,u,u,0,1,14,15,12,13,14,15]
-; AVX512DQ-FAST-NEXT: vpunpckhdq {{.*#+}} xmm7 = xmm10[2],xmm7[2],xmm10[3],xmm7[3]
+; AVX512DQ-FAST-NEXT: vmovdqa64 320(%rdi), %zmm26
+; AVX512DQ-FAST-NEXT: vmovdqa64 64(%rdi), %zmm1
+; AVX512DQ-FAST-NEXT: vbroadcasti32x4 {{.*#+}} ymm17 = [2,6,9,13,2,6,9,13]
+; AVX512DQ-FAST-NEXT: # ymm17 = mem[0,1,2,3,0,1,2,3]
+; AVX512DQ-FAST-NEXT: vmovdqa64 128(%rdi), %zmm23
+; AVX512DQ-FAST-NEXT: vbroadcasti32x4 {{.*#+}} ymm18 = [2,5,9,12,2,5,9,12]
+; AVX512DQ-FAST-NEXT: # ymm18 = mem[0,1,2,3,0,1,2,3]
+; AVX512DQ-FAST-NEXT: vbroadcasti32x4 {{.*#+}} ymm19 = [3,6,10,13,3,6,10,13]
+; AVX512DQ-FAST-NEXT: # ymm19 = mem[0,1,2,3,0,1,2,3]
+; AVX512DQ-FAST-NEXT: vmovdqa64 384(%rdi), %zmm20
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <2,6,9,u,13,u,u,u>
+; AVX512DQ-FAST-NEXT: vpermd %zmm1, %zmm0, %zmm7
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} ymm27 = <1,u,u,u,4,8,11,15>
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <2,5,9,u,12,u,u,u>
+; AVX512DQ-FAST-NEXT: vpermd %zmm1, %zmm0, %zmm10
+; AVX512DQ-FAST-NEXT: vpermd %zmm20, %zmm18, %zmm0
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <0,u,u,u,4,7,11,14>
+; AVX512DQ-FAST-NEXT: vpermd %zmm26, %zmm3, %zmm9
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [8,1,12,5,12,5,14,15]
+; AVX512DQ-FAST-NEXT: vpermd %zmm1, %zmm3, %zmm3
+; AVX512DQ-FAST-NEXT: vpermd %zmm23, %zmm19, %zmm4
+; AVX512DQ-FAST-NEXT: vmovdqa64 192(%rdi), %ymm25
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm25[0,1,0,2]
+; AVX512DQ-FAST-NEXT: vpbroadcastd {{.*#+}} ymm2 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
+; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm8, %ymm5
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm2, %ymm30
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[0,1,2,3,0,1,6,7,8,9,14,15,12,13,14,15,16,17,18,19,16,17,22,23,24,25,30,31,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm4[0,1,2,3,4,5,6],ymm5[7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[6,7,12,13,2,3,16,17,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FAST-NEXT: vmovdqa (%rdi), %ymm3
+; AVX512DQ-FAST-NEXT: vmovdqa 32(%rdi), %ymm4
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm3[0,1],ymm4[2],ymm3[3,4,5],ymm4[6],ymm3[7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm11, %xmm12
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2,3],xmm12[4],xmm11[5],xmm12[6],xmm11[7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm11[0,1,14,15,12,13,10,11,8,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm11[u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vporq %ymm6, %ymm11, %ymm22
+; AVX512DQ-FAST-NEXT: vpbroadcastw 252(%rdi), %xmm6
+; AVX512DQ-FAST-NEXT: vmovdqa 224(%rdi), %xmm13
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm13[u,u,u,u,u,u,u,u,0,1,14,15,12,13,14,15]
+; AVX512DQ-FAST-NEXT: vpunpckhdq {{.*#+}} xmm6 = xmm11[2],xmm6[2],xmm11[3],xmm6[3]
; AVX512DQ-FAST-NEXT: movw $992, %ax # imm = 0x3E0
; AVX512DQ-FAST-NEXT: kmovw %eax, %k1
-; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm7, %zmm6, %zmm22 {%k1}
+; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm6, %zmm5, %zmm22 {%k1}
; AVX512DQ-FAST-NEXT: vmovdqa 256(%rdi), %ymm5
-; AVX512DQ-FAST-NEXT: vmovdqa 288(%rdi), %ymm7
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm7[0,1],ymm5[2,3],ymm7[4,5],ymm5[6,7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm10, %xmm11
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3],xmm11[4],xmm10[5],xmm11[6,7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[u,u,u,u,u,u,10,11,8,9,6,7,4,5,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[0,1,6,7,8,9,14,15,8,9,14,15,4,5,2,3,16,17,22,23,24,25,30,31,24,25,30,31,20,21,18,19]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm8[0,1,2],xmm10[3,4,5,6],xmm8[7]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm8[4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpbroadcastq {{.*#+}} ymm6 = [16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31]
-; AVX512DQ-FAST-NEXT: vpshufb %ymm6, %ymm9, %ymm9
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm10[0,1,2,3,4,5],ymm9[6,7]
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm0, %zmm23
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[0,1,6,7,8,9,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm1[0,1,2],ymm4[3],ymm1[4,5],ymm4[6],ymm1[7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm9, %xmm10
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0],xmm9[1],xmm10[2,3,4,5],xmm9[6],xmm10[7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[2,3,0,1,14,15,12,13,10,11],zero,zero,zero,zero,zero,zero,zero,zero,ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vporq %ymm3, %ymm9, %ymm20
-; AVX512DQ-FAST-NEXT: vmovdqa 160(%rdi), %ymm3
-; AVX512DQ-FAST-NEXT: vmovdqa 128(%rdi), %ymm9
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm9[0,1],ymm3[2],ymm9[3,4,5],ymm3[6],ymm9[7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm10, %xmm11
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm10[0,1,2,3],xmm11[4],xmm10[5],xmm11[6],xmm10[7]
-; AVX512DQ-FAST-NEXT: vmovdqa 240(%rdi), %xmm10
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[0,1,0,1,14,15,12,13,10,11,8,9,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,22,23,28,29]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,6],ymm13[7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm13 = xmm10[0],xmm0[1],xmm10[2,3,4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[2,3,0,1,14,15,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm13, %zmm11, %zmm25
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm7[0,1,2],ymm5[3],ymm7[4,5],ymm5[6],ymm7[7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm11, %xmm13
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm13[0,1,2,3],xmm11[4],xmm13[5],xmm11[6],xmm13[7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[u,u,u,u,u,u,12,13,10,11,8,9,6,7,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[2,3,4,5,10,11,12,13,0,1,0,1,0,1,0,1,18,19,20,21,26,27,28,29,16,17,16,17,16,17,16,17]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm2[0,1,2],xmm11[3,4,5,6],xmm2[7]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm15[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27,28,29]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm11[6,7]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm2, %ymm16
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm12[2,3,4,5,10,11,16,17],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm4[0],ymm1[1],ymm4[2,3],ymm1[4],ymm4[5,6,7]
+; AVX512DQ-FAST-NEXT: vmovdqa 288(%rdi), %ymm6
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm11, %xmm12
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0],xmm12[1],xmm11[2,3,4,5],xmm12[6],xmm11[7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm11[4,5,2,3,0,1,14,15,12,13],zero,zero,zero,zero,zero,zero,zero,zero,ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %ymm2, %ymm11, %ymm11
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm12
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm12[0],xmm2[1],xmm12[2,3,4,5],xmm2[6],xmm12[7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[0,1,2,3,0,1,14,15,12,13,10,11,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512DQ-FAST-NEXT: vpbroadcastq {{.*#+}} ymm12 = [2,5,2,5,2,5,2,5]
-; AVX512DQ-FAST-NEXT: vpermd %ymm26, %ymm12, %ymm12
-; AVX512DQ-FAST-NEXT: vpshufhw {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm2[0,1,2,3,4,5,6],ymm12[7]
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm0[0],xmm10[0],xmm0[1],xmm10[1],xmm0[2],xmm10[2],xmm0[3],xmm10[3]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm0, %xmm31
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = [8,9,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX512DQ-FAST-NEXT: vpshufb %xmm0, %xmm2, %xmm15
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm0, %xmm18
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm15, %zmm12, %zmm15
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm5[0,1],ymm7[2],ymm5[3,4,5],ymm7[6],ymm5[7]
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm14
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm12, %xmm8
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm8 = xmm12[0,1,2,3],xmm8[4],xmm12[5],xmm8[6],xmm12[7]
-; AVX512DQ-FAST-NEXT: vpermd %zmm27, %zmm19, %zmm12
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[0,1,2,3,0,1,14,15,12,13,10,11,8,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm8[u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm12[2,3,16,17,22,23,24,25,30,31,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %ymm12, %ymm8, %ymm8
-; AVX512DQ-FAST-NEXT: vpermd %zmm17, %zmm21, %zmm12
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm17 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm11, %zmm17, %zmm15
-; AVX512DQ-FAST-NEXT: vpshufb %ymm6, %ymm12, %ymm11
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm11[6,7]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm12[0,1,2],xmm11[3],xmm12[4],xmm11[5],xmm12[6,7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[u,u,u,u,u,u,10,11,8,9,6,7,4,5,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[0,1,6,7,8,9,14,15,8,9,14,15,4,5,2,3,16,17,22,23,24,25,30,31,24,25,30,31,20,21,18,19]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm9[0,1,2],xmm11[3,4,5,6],xmm9[7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm9[4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpbroadcastq {{.*#+}} ymm9 = [16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31]
+; AVX512DQ-FAST-NEXT: vpshufb %ymm9, %ymm0, %ymm0
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1,2,3,4,5],ymm0[6,7]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm0, %ymm28
+; AVX512DQ-FAST-NEXT: vmovdqa 240(%rdi), %xmm15
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2],ymm4[3],ymm3[4,5],ymm4[6],ymm3[7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm0, %xmm11
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm11[0],xmm0[1],xmm11[2,3,4,5],xmm0[6],xmm11[7]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = <2,3,0,1,14,15,12,13,10,11,128,128,128,128,128,128,128,128,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512DQ-FAST-NEXT: vpshufb %ymm11, %ymm0, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[0,1,6,7,8,9,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FAST-NEXT: vporq %ymm10, %ymm0, %ymm21
+; AVX512DQ-FAST-NEXT: vmovdqa 160(%rdi), %ymm12
+; AVX512DQ-FAST-NEXT: vmovdqa 128(%rdi), %ymm10
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1],ymm12[2],ymm10[3,4,5],ymm12[6],ymm10[7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm0, %xmm14
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm14[4],xmm0[5],xmm14[6],xmm0[7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,14,15,12,13,10,11,8,9,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,22,23,28,29]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm8[7]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm8 = xmm15[0],xmm13[1],xmm15[2,3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm11, %xmm8, %xmm8
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm0, %zmm24
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[2,3,4,5,10,11,16,17],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6,7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm7, %xmm8
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0],xmm8[1],xmm7[2,3,4,5],xmm8[6],xmm7[7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm7 = ymm7[4,5,2,3,0,1,14,15,12,13],zero,zero,zero,zero,zero,zero,zero,zero,ymm7[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %ymm0, %ymm7, %ymm8
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm12[3],ymm10[4,5],ymm12[6],ymm10[7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm0, %xmm7
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0],xmm0[1],xmm7[2,3,4,5],xmm0[6],xmm7[7]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <0,1,2,3,0,1,14,15,12,13,10,11,8,9,128,128,128,128,128,128,128,128,128,128,u,u,u,u,u,u,u,u>
+; AVX512DQ-FAST-NEXT: vpshufb %xmm0, %xmm7, %xmm7
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
+; AVX512DQ-FAST-NEXT: vpbroadcastq {{.*#+}} ymm11 = [2,5,2,5,2,5,2,5]
+; AVX512DQ-FAST-NEXT: vpermd %ymm25, %ymm11, %ymm11
+; AVX512DQ-FAST-NEXT: vpshufhw {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm7[0,1,2,3,4,5,6],ymm11[7]
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm13[0],xmm15[0],xmm13[1],xmm15[1],xmm13[2],xmm15[2],xmm13[3],xmm15[3]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm14 = [8,9,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm14, %xmm7, %xmm2
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm14, %xmm31
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm11, %zmm16
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2],ymm5[3],ymm6[4,5],ymm5[6],ymm6[7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm11
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm11[0,1,2,3],xmm2[4],xmm11[5],xmm2[6],xmm11[7]
+; AVX512DQ-FAST-NEXT: vpermd %zmm26, %zmm27, %zmm11
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,12,13,10,11,8,9,6,7,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm11[2,3,4,5,10,11,12,13,0,1,0,1,0,1,0,1,18,19,20,21,26,27,28,29,16,17,16,17,16,17,16,17]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm11[0,1,2],xmm2[3,4,5,6],xmm11[7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpermd %zmm20, %zmm17, %zmm11
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27,28,29]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm11[6,7]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm2, %ymm29
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm6[2],ymm5[3,4,5],ymm6[6],ymm5[7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm14
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm14[4],xmm2[5],xmm14[6],xmm2[7]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = [10,3,6,15,12,13,6,15]
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} ymm27 = <1,u,u,u,5,8,12,15>
+; AVX512DQ-FAST-NEXT: vpshufb %ymm0, %ymm2, %ymm0
+; AVX512DQ-FAST-NEXT: vpermd %zmm26, %zmm27, %zmm2
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[2,3,16,17,22,23,24,25,30,31,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %ymm2, %ymm0, %ymm0
+; AVX512DQ-FAST-NEXT: vpermd %zmm20, %zmm19, %zmm2
+; AVX512DQ-FAST-NEXT: vpshufb %ymm9, %ymm2, %ymm2
+; AVX512DQ-FAST-NEXT: vpermd %zmm1, %zmm14, %zmm9
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm19 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm8, %zmm19, %zmm16
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX512DQ-FAST-NEXT: movw $-512, %ax # imm = 0xFE00
; AVX512DQ-FAST-NEXT: kmovw %eax, %k1
-; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm8, %zmm0, %zmm15 {%k1}
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm8 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[4,5,10,11,0,1,22,23],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm4[0],ymm1[1],ymm4[2,3,4],ymm1[5],ymm4[6,7]
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm4, %ymm6
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm1, %ymm4
+; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm16 {%k1}
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm9[4,5,10,11,0,1,22,23],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0],ymm3[1],ymm4[2,3,4],ymm3[5],ymm4[6,7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm8
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm8[0],xmm2[1],xmm8[2],xmm2[3],xmm8[4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[6,7,4,5,2,3,0,1,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %ymm0, %ymm2, %ymm8
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0],ymm10[1],ymm12[2,3],ymm10[4],ymm12[5,6,7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm0, %xmm2
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3,4,5],xmm2[6],xmm0[7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,2,3,0,1,14,15,12,13,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm2
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm25[0,1,1,3]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm9 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,24,25]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm9[7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[12,13,10,11,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm2, %zmm20
+; AVX512DQ-FAST-NEXT: vmovdqa 416(%rdi), %ymm2
+; AVX512DQ-FAST-NEXT: vmovdqa 384(%rdi), %ymm9
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm9[0,1],ymm2[2],ymm9[3,4,5],ymm2[6],ymm9[7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm7, %xmm14
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3],xmm14[4],xmm7[5],xmm14[6],xmm7[7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm5[0,1,2],ymm6[3],ymm5[4,5],ymm6[6],ymm5[7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm14, %xmm11
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0],xmm14[1],xmm11[2,3,4,5],xmm14[6],xmm11[7]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = [2,11,2,11,12,5,8,9]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,0,1,14,15,12,13,10,11,8,9]
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
+; AVX512DQ-FAST-NEXT: vpermd %zmm26, %zmm14, %zmm14
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm14 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm14[0,1,22,23,28,29,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm14[0,1,2],ymm7[3,4,5,6,7],ymm14[8,9,10],ymm7[11,12,13,14,15]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[0,1,2,3,2,3,0,1,14,15,12,13,10,11],zero,zero
+; AVX512DQ-FAST-NEXT: vpor %ymm14, %ymm11, %ymm11
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2,3],ymm7[4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm8, %zmm19, %zmm20
+; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm7, %zmm0, %zmm20 {%k1}
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = [8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm7, %xmm8, %xmm11
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm8, %xmm8
+; AVX512DQ-FAST-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm11[0],xmm8[0],xmm11[1],xmm8[1],xmm11[2],xmm8[2],xmm11[3],xmm8[3]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm8 = <0,3,7,10,14,u,u,u>
+; AVX512DQ-FAST-NEXT: vpermd %zmm1, %zmm8, %zmm14
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm8 = <u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31>
+; AVX512DQ-FAST-NEXT: vpshufb %ymm8, %ymm14, %ymm14
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1],ymm14[2,3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm12[0],ymm10[1],ymm12[2,3,4],ymm10[5],ymm12[6,7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm10, %xmm12
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm12[0],xmm10[1],xmm12[2],xmm10[3],xmm12[4,5,6,7]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm30, %ymm12
+; AVX512DQ-FAST-NEXT: vpshufb %ymm12, %ymm0, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[0,1,6,7,4,5,2,3,0,1,14,15,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3,4,5,6],ymm0[7]
+; AVX512DQ-FAST-NEXT: vpbroadcastw 232(%rdi), %xmm10
+; AVX512DQ-FAST-NEXT: vpsrlq $48, %xmm15, %xmm12
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm10[0],xmm12[0],xmm10[1],xmm12[1],xmm10[2],xmm12[2],xmm10[3],xmm12[3]
+; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm10, %zmm0, %zmm10
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm11, %zmm19, %zmm10
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2],ymm2[3],ymm9[4,5],ymm2[6],ymm9[7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm0, %xmm11
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm11[0],xmm0[1],xmm11[2,3,4,5],xmm0[6],xmm11[7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm6[0],ymm5[1],ymm6[2,3],ymm5[4],ymm6[5,6,7]
; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm11, %xmm12
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm12[0],xmm11[1],xmm12[2],xmm11[3],xmm12[4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm11[6,7,4,5,2,3,0,1,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %ymm8, %ymm11, %ymm12
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm3[0],ymm9[1],ymm3[2,3],ymm9[4],ymm3[5,6,7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm8, %xmm11
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0],xmm11[1],xmm8[2,3,4,5],xmm11[6],xmm8[7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[0,1,4,5,2,3,0,1,14,15,12,13,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm11 = ymm26[0,1,1,3]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm13 = ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,24,25]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm13[7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[12,13,10,11,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm8, %zmm19
-; AVX512DQ-FAST-NEXT: vmovdqa 416(%rdi), %ymm8
-; AVX512DQ-FAST-NEXT: vmovdqa 384(%rdi), %ymm5
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm5[0,1],ymm8[2],ymm5[3,4,5],ymm8[6],ymm5[7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm13, %xmm0
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm13[0,1,2,3],xmm0[4],xmm13[5],xmm0[6],xmm13[7]
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm14, %ymm2
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2],ymm7[3],ymm14[4,5],ymm7[6],ymm14[7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm13, %xmm14
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm13 = xmm14[0],xmm13[1],xmm14[2,3,4,5],xmm13[6],xmm14[7]
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <0,3,7,10,14,u,u,u>
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} ymm28 = [2,11,2,11,12,5,8,9]
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm16, %zmm0, %zmm21
-; AVX512DQ-FAST-NEXT: vpermd %zmm27, %zmm28, %zmm1
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm12, %zmm17, %zmm19
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,0,1,14,15,12,13,10,11,8,9]
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0,1,22,23,28,29,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm13[0,1,2,3,2,3,0,1,14,15,12,13,10,11],zero,zero
-; AVX512DQ-FAST-NEXT: vpor %ymm1, %ymm12, %ymm1
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm19 {%k1}
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1],ymm4[2],ymm6[3,4],ymm4[5],ymm6[6,7]
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm13 = [8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX512DQ-FAST-NEXT: vpshufb %xmm13, %xmm0, %xmm1
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX512DQ-FAST-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX512DQ-FAST-NEXT: vpermd %zmm29, %zmm14, %zmm1
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm12 = <u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31>
-; AVX512DQ-FAST-NEXT: vpshufb %ymm12, %ymm1, %ymm1
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm12, %ymm28
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0],ymm9[1],ymm3[2,3,4],ymm9[5],ymm3[6,7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm1, %xmm3
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2],xmm1[3],xmm3[4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,1,6,7,4,5,2,3,0,1,14,15,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7]
-; AVX512DQ-FAST-NEXT: vpbroadcastw 232(%rdi), %xmm3
-; AVX512DQ-FAST-NEXT: vpsrlq $48, %xmm10, %xmm9
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3]
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = <2,u,u,u,6,9,13,u>
-; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm3, %zmm1, %zmm3
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm0, %zmm17, %zmm3
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm8[3],ymm5[4,5],ymm8[6],ymm5[7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3,4,5],xmm0[6],xmm1[7]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0],ymm2[1],ymm7[2,3],ymm2[4],ymm7[5,6,7]
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm2, %ymm12
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm1, %xmm11
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm11[1],xmm1[2,3,4,5],xmm11[6],xmm1[7]
-; AVX512DQ-FAST-NEXT: vpermd %zmm27, %zmm9, %zmm9
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0],xmm12[1],xmm11[2,3,4,5],xmm12[6],xmm11[7]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm12 = <2,u,u,u,6,9,13,u>
+; AVX512DQ-FAST-NEXT: vpermd %zmm26, %zmm12, %zmm12
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,2,3,0,1,14,15,12,13,10,11]
; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm9 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm9[2,3,16,17,22,23,24,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm9[0,1,2],ymm0[3,4,5,6,7],ymm9[8,9,10],ymm0[11,12,13,14,15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,2,3,0,1,14,15,12,13],zero,zero
-; AVX512DQ-FAST-NEXT: vpor %ymm1, %ymm9, %ymm1
-; AVX512DQ-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [2,6,9,13,2,6,9,13]
-; AVX512DQ-FAST-NEXT: # ymm2 = mem[0,1,0,1]
-; AVX512DQ-FAST-NEXT: vpermd %zmm24, %zmm2, %zmm11
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = <0,4,7,11,14,u,u,u>
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpermd %zmm24, %zmm30, %zmm1
-; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm3 {%k1}
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <0,3,3,u,0,3,7,u>
-; AVX512DQ-FAST-NEXT: vpermd %ymm26, %ymm0, %ymm0
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,0,1,6,7,8,9,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[0,1,0,1,6,7,8,9,14,15,u,u,u,u,u,u,16,17,16,17,22,23,24,25,30,31,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7],ymm1[8,9,10,11,12],ymm0[13,14,15]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm31, %xmm2
-; AVX512DQ-FAST-NEXT: vpsrld $16, %xmm31, %xmm1
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
-; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm1, %zmm0, %zmm0
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm4[2,3],ymm6[4,5],ymm4[6,7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm1, %xmm14
-; AVX512DQ-FAST-NEXT: vpshufb %xmm13, %xmm14, %xmm13
-; AVX512DQ-FAST-NEXT: vpermd %zmm29, %zmm9, %zmm9
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[10,11,6,7,4,5,6,7,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,u,u,u,u,u,2,3,4,5,10,11,12,13,18,19,18,19,18,19,18,19,18,19,20,21,26,27,28,29]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm1[0,1],ymm9[2,3,4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0],ymm5[1],ymm8[2,3],ymm5[4],ymm8[5,6,7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm1, %xmm13
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm13[1],xmm1[2,3,4,5],xmm13[6],xmm1[7]
-; AVX512DQ-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm13 = [0,4,7,0,0,4,7,0]
-; AVX512DQ-FAST-NEXT: # ymm13 = mem[0,1,0,1]
-; AVX512DQ-FAST-NEXT: vpermd %ymm26, %ymm13, %ymm13
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,u,u,u,u,u,u,u,2,3,4,5,10,11,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm11[4,5,2,3,4,5,10,11,12,13,u,u,u,u,u,u,20,21,18,19,20,21,26,27,28,29,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm13[5,6,7],ymm11[8,9,10,11,12],ymm13[13,14,15]
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm10 = xmm2[4],xmm10[4],xmm2[5],xmm10[5],xmm2[6],xmm10[6],xmm2[7],xmm10[7]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm18, %xmm2
-; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm10, %xmm10
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = <3,u,u,u,6,10,13,u>
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm10, %zmm11, %zmm10
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm7[0],ymm12[1],ymm7[2,3,4],ymm12[5],ymm7[6,7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm11, %xmm14
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm11 = xmm14[0],xmm11[1],xmm14[2],xmm11[3],xmm14[4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpermd %zmm27, %zmm13, %zmm13
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,4,5,2,3,0,1,14,15,12,13]
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm13 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm13[0,1,18,19,20,21,26,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm13[0,1,2],ymm1[3,4,5,6,7],ymm13[8,9,10],ymm1[11,12,13,14,15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[0,1,2,3,6,7,4,5,2,3,0,1,14,15],zero,zero
-; AVX512DQ-FAST-NEXT: vpor %ymm13, %ymm11, %ymm11
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1,2,3],ymm1[4,5,6,7]
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = <1,4,8,11,15,u,u,u>
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm0, %zmm17, %zmm9
-; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm1, %zmm0, %zmm9 {%k1}
-; AVX512DQ-FAST-NEXT: vpermd %zmm29, %zmm11, %zmm0
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm28, %ymm1
-; AVX512DQ-FAST-NEXT: vpshufb %ymm1, %ymm0, %ymm0
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1,2],ymm4[3],ymm6[4,5],ymm4[6],ymm6[7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm1, %xmm4
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,u,u,10,11,6,7,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm12[2,3,16,17,22,23,24,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm12[0,1,2],ymm0[3,4,5,6,7],ymm12[8,9,10],ymm0[11,12,13,14,15]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5,2,3,0,1,14,15,12,13],zero,zero
+; AVX512DQ-FAST-NEXT: vpor %ymm12, %ymm11, %ymm11
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm0[4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpermd %zmm23, %zmm17, %zmm0
+; AVX512DQ-FAST-NEXT: vpermd %zmm23, %zmm18, %zmm12
+; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm11, %zmm0, %zmm10 {%k1}
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = <0,3,3,u,0,3,7,u>
+; AVX512DQ-FAST-NEXT: vpermd %ymm25, %ymm11, %ymm11
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,u,u,u,u,u,u,0,1,6,7,8,9,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm12 = ymm12[0,1,0,1,6,7,8,9,14,15,u,u,u,u,u,u,16,17,16,17,22,23,24,25,30,31,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm12[0,1,2,3,4],ymm11[5,6,7],ymm12[8,9,10,11,12],ymm11[13,14,15]
+; AVX512DQ-FAST-NEXT: vpsrld $16, %xmm13, %xmm12
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm12 = xmm12[4],xmm15[4],xmm12[5],xmm15[5],xmm12[6],xmm15[6],xmm12[7],xmm15[7]
+; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm12, %zmm11, %zmm11
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm4[0,1],ymm3[2,3],ymm4[4,5],ymm3[6,7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm12, %xmm14
+; AVX512DQ-FAST-NEXT: vpshufb %xmm7, %xmm14, %xmm7
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <0,4,7,11,14,u,u,u>
+; AVX512DQ-FAST-NEXT: vpermd %zmm1, %zmm14, %zmm14
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm12[10,11,6,7,4,5,6,7,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm12[0],xmm7[0],xmm12[1],xmm7[1],xmm12[2],xmm7[2],xmm12[3],xmm7[3]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm12 = ymm14[u,u,u,u,u,u,u,u,2,3,4,5,10,11,12,13,18,19,18,19,18,19,18,19,18,19,20,21,26,27,28,29]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm7[0,1],ymm12[2,3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm2[0],ymm9[1],ymm2[2,3],ymm9[4],ymm2[5,6,7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm7, %xmm14
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0],xmm14[1],xmm7[2,3,4,5],xmm14[6],xmm7[7]
+; AVX512DQ-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm14 = [0,4,7,0,0,4,7,0]
+; AVX512DQ-FAST-NEXT: # ymm14 = mem[0,1,0,1]
+; AVX512DQ-FAST-NEXT: vpermd %ymm25, %ymm14, %ymm14
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm14 = ymm14[u,u,u,u,u,u,u,u,u,u,2,3,4,5,10,11,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[4,5,2,3,4,5,10,11,12,13,u,u,u,u,u,u,20,21,18,19,20,21,26,27,28,29,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7],ymm0[8,9,10,11,12],ymm14[13,14,15]
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm13 = xmm13[4],xmm15[4],xmm13[5],xmm15[5],xmm13[6],xmm15[6],xmm13[7],xmm15[7]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm31, %xmm14
+; AVX512DQ-FAST-NEXT: vpshufb %xmm14, %xmm13, %xmm13
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm13, %zmm0, %zmm13
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0],ymm5[1],ymm6[2,3,4],ymm5[5],ymm6[6,7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm0, %xmm14
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm14[0],xmm0[1],xmm14[2],xmm0[3],xmm14[4,5,6,7]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <3,u,u,u,6,10,13,u>
+; AVX512DQ-FAST-NEXT: vpermd %zmm26, %zmm14, %zmm14
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,4,5,2,3,0,1,14,15,12,13]
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm14 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm14[0,1,18,19,20,21,26,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm14[0,1,2],ymm7[3,4,5,6,7],ymm14[8,9,10],ymm7[11,12,13,14,15]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,4,5,2,3,0,1,14,15],zero,zero
+; AVX512DQ-FAST-NEXT: vpor %ymm0, %ymm14, %ymm0
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm11, %zmm19, %zmm12
+; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm12 {%k1}
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <1,4,8,11,15,u,u,u>
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm28, %zmm0, %zmm7
+; AVX512DQ-FAST-NEXT: vpermd %zmm1, %zmm0, %zmm0
+; AVX512DQ-FAST-NEXT: vpshufb %ymm8, %ymm0, %ymm0
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm3[3],ymm4[4,5],ymm3[6],ymm4[7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm1, %xmm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,u,10,11,6,7,4,5,6,7]
; AVX512DQ-FAST-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,4,6,7]
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0],ymm5[1],ymm8[2,3,4],ymm5[5],ymm8[6,7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm0, %xmm2
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2],xmm0[3],xmm2[4,5,6,7]
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [0,1,10,3,14,7,10,3]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0,1],ymm12[2],ymm7[3,4],ymm12[5],ymm7[6,7]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX512DQ-FAST-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,3,1,3,4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[8,9,8,9,4,5,6,7,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
-; AVX512DQ-FAST-NEXT: vpermd %zmm27, %zmm2, %zmm2
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[4,5,10,11,u,u,u,u,u,u,u,u,0,1,14,15,20,21,26,27,16,17,26,27,16,17,20,21,16,17,30,31]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm4[1,2],ymm2[3,4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,6,7,4,5,2,3,0,1,14,15]
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7],ymm2[8,9,10],ymm0[11,12,13,14,15]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm10, %zmm17, %zmm1
-; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm1 {%k1}
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm9[1],ymm2[2,3,4],ymm9[5],ymm2[6,7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3],xmm2[4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1],ymm5[2],ymm6[3,4],ymm5[5],ymm6[6,7]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512DQ-FAST-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,3,1,3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[8,9,8,9,4,5,6,7,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,10,3,14,7,10,3]
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm29, %zmm0, %zmm4
+; AVX512DQ-FAST-NEXT: vpermd %zmm26, %zmm3, %zmm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[4,5,10,11,u,u,u,u,u,u,u,u,0,1,14,15,20,21,26,27,16,17,26,27,16,17,20,21,16,17,30,31]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0],ymm2[1,2],ymm3[3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,6,7,4,5,2,3,0,1,14,15]
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7],ymm2[8,9,10],ymm1[11,12,13,14,15]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm13, %zmm19, %zmm0
+; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm1, %zmm0, %zmm0 {%k1}
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm22, %zmm1, %zmm7
+; AVX512DQ-FAST-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm21, %zmm24
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm24, %zmm1, %zmm4
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm7, (%rsi)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm4, (%rdx)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm16, (%rcx)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm20, (%r8)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm10, (%r9)
; AVX512DQ-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm22, %zmm0, %zmm23
-; AVX512DQ-FAST-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm20, %zmm25
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm25, %zmm0, %zmm21
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm23, (%rsi)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm21, (%rdx)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm15, (%rcx)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm19, (%r8)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm3, (%r9)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm9, (%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm12, (%rax)
; AVX512DQ-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm1, (%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm0, (%rax)
; AVX512DQ-FAST-NEXT: vzeroupper
; AVX512DQ-FAST-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
index a08234ffb482f5..40355e3f7686f6 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
@@ -305,56 +305,48 @@ define void @load_i32_stride4_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-LABEL: load_i32_stride4_vf8:
; AVX2-ONLY: # %bb.0:
; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm3 = [0,4,0,4,0,4,0,4]
-; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm3, %ymm5
-; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm3, %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = [0,4,0,4]
-; AVX2-ONLY-NEXT: # xmm5 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm5, %ymm6
-; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm7
-; AVX2-ONLY-NEXT: vmovaps 16(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm5
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm9 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm6 = [1,5,1,5,1,5,1,5]
-; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm6, %ymm9
-; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm6, %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vmovaps 48(%rdi), %xmm9
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm10 = xmm5[0],xmm9[0],xmm5[1],xmm9[1]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [1,5,1,5]
-; AVX2-ONLY-NEXT: # xmm11 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm0, %ymm11, %ymm11
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm11[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm10 = [2,6,2,6,2,6,2,6]
-; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm10, %ymm11
-; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm10, %ymm10
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [2,6,2,6]
-; AVX2-ONLY-NEXT: # xmm11 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm11, %ymm4
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm7 = xmm7[2],xmm8[2],xmm7[3],xmm8[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm7 = [3,7,3,7,3,7,3,7]
-; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm7, %ymm2
-; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm7, %ymm1
+; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm4 = [0,4,0,4,0,4,0,4]
+; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm4, %ymm5
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm4, %ymm6
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm4, %ymm4
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm6
+; AVX2-ONLY-NEXT: vmovaps 16(%rdi), %xmm7
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm8
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm9 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm9[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm5 = [1,5,1,5,1,5,1,5]
+; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm5, %ymm9
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm5, %ymm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vmovaps 48(%rdi), %xmm10
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm11 = xmm8[0],xmm10[0],xmm8[1],xmm10[1]
+; AVX2-ONLY-NEXT: vpermps %ymm0, %ymm5, %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm11[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm9 = [2,6,2,6,2,6,2,6]
+; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm9, %ymm11
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm9, %ymm12
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm9, %ymm3
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm6 = xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm6[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm6 = [3,7,3,7,3,7,3,7]
+; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm6, %ymm2
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm6, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm2 = xmm5[2],xmm9[2],xmm5[3],xmm9[3]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = [3,7,3,7]
-; AVX2-ONLY-NEXT: # xmm5 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm0, %ymm5, %ymm0
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm2 = xmm8[2],xmm10[2],xmm8[3],xmm10[3]
+; AVX2-ONLY-NEXT: vpermps %ymm0, %ymm6, %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovaps %ymm3, (%rsi)
-; AVX2-ONLY-NEXT: vmovaps %ymm6, (%rdx)
-; AVX2-ONLY-NEXT: vmovaps %ymm4, (%rcx)
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%rsi)
+; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rdx)
+; AVX2-ONLY-NEXT: vmovaps %ymm3, (%rcx)
; AVX2-ONLY-NEXT: vmovaps %ymm0, (%r8)
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
@@ -634,38 +626,36 @@ define void @load_i32_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-LABEL: load_i32_stride4_vf16:
; AVX2-ONLY: # %bb.0:
; AVX2-ONLY-NEXT: subq $104, %rsp
-; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm8
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm8
; AVX2-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm7
-; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm4
; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm7
; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm2
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm0 = [0,4,0,4,0,4,0,4]
; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm0, %ymm6
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm0, %ymm6
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovaps 144(%rdi), %xmm10
; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm11
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = [0,4,0,4]
-; AVX2-ONLY-NEXT: # xmm9 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm12, %ymm9, %ymm12
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm12[2,3]
+; AVX2-ONLY-NEXT: vpermps %ymm9, %ymm0, %ymm9
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm9[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm9, %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm0, %ymm6
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm12
; AVX2-ONLY-NEXT: vmovaps 16(%rdi), %xmm13
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm13[0],xmm12[1],xmm13[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm6[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm0 = [1,5,1,5,1,5,1,5]
; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm0, %ymm1
@@ -674,68 +664,62 @@ define void @load_i32_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm14
; AVX2-ONLY-NEXT: vmovaps 48(%rdi), %xmm6
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm9 = xmm14[0],xmm6[0],xmm14[1],xmm6[1]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [1,5,1,5]
-; AVX2-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm7, %ymm15
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm0, %ymm15
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm15[0,1],xmm9[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm0, %ymm9
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm7, %ymm7
-; AVX2-ONLY-NEXT: vmovaps 176(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm8 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm7 = [2,6,2,6,2,6,2,6]
-; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm7, %ymm8
-; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm7, %ymm9
+; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm0, %ymm0
+; AVX2-ONLY-NEXT: vmovaps 176(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm8 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm8[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm0 = [2,6,2,6,2,6,2,6]
+; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm0, %ymm8
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm0, %ymm9
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6,7]
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm9 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm10 = [2,6,2,6]
-; AVX2-ONLY-NEXT: # xmm10 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm11 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0,1],xmm11[2,3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0,1],xmm10[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm7, %ymm9
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm7, %ymm7
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm0, %ymm9
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm0, %ymm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm10 = xmm12[2],xmm13[2],xmm12[3],xmm13[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm10[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm9 = [3,7,3,7,3,7,3,7]
; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm9, %ymm5
; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm9, %ymm4
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm5 = xmm14[2],xmm6[2],xmm14[3],xmm6[3]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = [3,7,3,7]
-; AVX2-ONLY-NEXT: # xmm6 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm9, %ymm2
-; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm9, %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm6, %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, (%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
-; AVX2-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, (%rdx)
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm9, %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm9, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rdx)
+; AVX2-ONLY-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rdx)
; AVX2-ONLY-NEXT: vmovaps %ymm8, 32(%rcx)
-; AVX2-ONLY-NEXT: vmovaps %ymm7, (%rcx)
-; AVX2-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm0, (%rcx)
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%r8)
; AVX2-ONLY-NEXT: vmovaps %ymm4, (%r8)
; AVX2-ONLY-NEXT: addq $104, %rsp
; AVX2-ONLY-NEXT: vzeroupper
@@ -1337,243 +1321,228 @@ define void @load_i32_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-LABEL: load_i32_stride4_vf32:
; AVX2-ONLY: # %bb.0:
; AVX2-ONLY-NEXT: subq $680, %rsp # imm = 0x2A8
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm6
; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %ymm8
; AVX2-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %ymm10
+; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm11
+; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm9
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm0 = [0,4,0,4,0,4,0,4]
-; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vmovaps %ymm2, %ymm7
-; AVX2-ONLY-NEXT: vpermps %ymm11, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vmovaps 272(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vpermps %ymm9, %ymm0, %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vmovaps %ymm3, %ymm7
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vmovaps 272(%rdi), %xmm3
; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [0,4,0,4]
-; AVX2-ONLY-NEXT: # xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm1, %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX2-ONLY-NEXT: vpermps %ymm11, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm0, %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vmovaps 400(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vpermps %ymm6, %ymm0, %ymm2
; AVX2-ONLY-NEXT: vmovaps %ymm6, %ymm12
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm0, %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vmovaps 400(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm1, %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovaps %ymm3, %ymm13
-; AVX2-ONLY-NEXT: vpermps %ymm9, %ymm0, %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm5
; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 144(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps 144(%rdi), %xmm3
; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm1, %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm14
; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm0, %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm1, %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm3
; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 16(%rdi), %xmm2
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm1 = [1,5,1,5,1,5,1,5]
-; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm1, %ymm0
-; AVX2-ONLY-NEXT: vmovaps %ymm7, %ymm8
-; AVX2-ONLY-NEXT: vpermps %ymm11, %ymm1, %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 304(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm0 = [1,5,1,5,1,5,1,5]
+; AVX2-ONLY-NEXT: vpermps %ymm9, %ymm0, %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vmovaps %ymm7, %ymm6
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm5
+; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 304(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [1,5,1,5]
-; AVX2-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm7, %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm1, %ymm0
-; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm1, %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm2
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 48(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
-; AVX2-ONLY-NEXT: vpermps %ymm0, %ymm7, %ymm10
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm10[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm13, %ymm1, %ymm5
-; AVX2-ONLY-NEXT: vmovaps %ymm13, %ymm10
-; AVX2-ONLY-NEXT: vpermps %ymm9, %ymm1, %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 176(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm7, %ymm13
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm13[0,1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps %ymm12, %ymm3
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm0, %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm7
+; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps 48(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm5 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm0, %ymm7
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm0, %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm12, %ymm0, %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm12
; AVX2-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm12, %ymm1, %ymm0
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm1, %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovaps 176(%rdi), %xmm7
+; AVX2-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm5[0],xmm7[0],xmm5[1],xmm7[1]
+; AVX2-ONLY-NEXT: vpermps %ymm12, %ymm0, %ymm13
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm13[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm0, %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm0, %ymm7
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm7[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm7, %ymm13
-; AVX2-ONLY-NEXT: vmovaps 432(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %xmm5
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm12 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm13[0,1],xmm12[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm0, %ymm0
+; AVX2-ONLY-NEXT: vmovaps 432(%rdi), %xmm7
+; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm12 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm12[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm0 = [2,6,2,6,2,6,2,6]
-; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm0, %ymm12
-; AVX2-ONLY-NEXT: vmovaps %ymm11, %ymm7
-; AVX2-ONLY-NEXT: vpermps %ymm11, %ymm0, %ymm13
+; AVX2-ONLY-NEXT: vpermps %ymm9, %ymm0, %ymm12
+; AVX2-ONLY-NEXT: vpermps %ymm6, %ymm0, %ymm13
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm13 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm13 = xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [2,6,2,6]
-; AVX2-ONLY-NEXT: # xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm13 = xmm13[2],mem[2],xmm13[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps %ymm11, %ymm0, %ymm11
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm13[0,1],xmm11[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm0, %ymm11
-; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm0, %ymm12
+; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm0, %ymm11
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm0, %ymm12
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm12 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm12 = xmm2[2],mem[2],xmm2[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovaps (%rsp), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm12 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm12 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm12[0,1],xmm13[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm0, %ymm11
-; AVX2-ONLY-NEXT: vmovaps %ymm9, %ymm3
-; AVX2-ONLY-NEXT: vpermps %ymm9, %ymm0, %ymm12
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm0, %ymm11
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm0, %ymm12
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm12 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm12 = xmm2[2],mem[2],xmm2[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm12 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm12 = xmm10[2],mem[2],xmm10[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm12[0,1],xmm13[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm10, (%rsp) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm0, %ymm11
-; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm11 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm11 = xmm2[2],mem[2],xmm2[3],mem[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm11[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm1 = [3,7,3,7,3,7,3,7]
-; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm1, %ymm0
-; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm1, %ymm11
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm2 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [3,7,3,7]
-; AVX2-ONLY-NEXT: # xmm4 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm9 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm9[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm1, %ymm2
-; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm1, %ymm9
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vmovaps (%rsp), %xmm7 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm9 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm9 = xmm7[2],mem[2],xmm7[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm11 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm11[0,1],xmm9[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm1, %ymm8
-; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm1, %ymm7
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm8 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm8 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm9 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm3 = xmm5[2],xmm6[2],xmm5[3],xmm6[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 32(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 96(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 64(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, (%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 96(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 32(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, (%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 64(%rdx)
-; AVX2-ONLY-NEXT: vmovaps %ymm12, 32(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 96(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 64(%rcx)
+; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm0, %ymm12
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm12 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm12 = xmm10[2],mem[2],xmm10[3],mem[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm12[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm0 = [3,7,3,7,3,7,3,7]
+; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm0, %ymm11
+; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm0, %ymm12
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm2 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm9, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vpermps %ymm6, %ymm0, %ymm6
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm0, %ymm6
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm0, %ymm8
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm8[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 96(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 64(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, (%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 96(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, (%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 64(%rdx)
+; AVX2-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 96(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 64(%rcx)
; AVX2-ONLY-NEXT: vmovaps %ymm13, (%rcx)
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 96(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm7, 32(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 64(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm0, (%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm0, 96(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 64(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm2, (%r8)
; AVX2-ONLY-NEXT: addq $680, %rsp # imm = 0x2A8
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
@@ -2812,497 +2781,494 @@ define void @load_i32_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX2-ONLY-LABEL: load_i32_stride4_vf64:
; AVX2-ONLY: # %bb.0:
-; AVX2-ONLY-NEXT: subq $1992, %rsp # imm = 0x7C8
+; AVX2-ONLY-NEXT: subq $1944, %rsp # imm = 0x798
+; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm7
; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %ymm8
; AVX2-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm14
+; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm4
; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm10
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm10
+; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm14
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm2 = [0,4,0,4,0,4,0,4]
-; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm2, %ymm1
-; AVX2-ONLY-NEXT: vmovaps %ymm3, %ymm13
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm2, %ymm0
+; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm2, %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovaps 144(%rdi), %xmm3
; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [0,4,0,4]
-; AVX2-ONLY-NEXT: # xmm3 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm3, %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vpermps %ymm9, %ymm2, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm2, %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm1
+; AVX2-ONLY-NEXT: vmovaps %ymm4, %ymm9
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 400(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 400(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
-; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm3, %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm12
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm2, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %ymm3
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm0, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vpermps %ymm12, %ymm2, %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm2, %ymm0
+; AVX2-ONLY-NEXT: vmovaps %ymm3, %ymm11
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm2, %ymm1
+; AVX2-ONLY-NEXT: vmovaps %ymm7, %ymm13
+; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 656(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 656(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm3, %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vpermps %ymm11, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm1
+; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm2, %ymm0
+; AVX2-ONLY-NEXT: vmovaps %ymm1, %ymm8
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm2, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 912(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 912(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm3, %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermps %ymm0, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm2, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 272(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 272(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm3, %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 608(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermps %ymm0, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm2, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 528(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 528(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm3, %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermps %ymm0, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm2, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 784(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 784(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm3, %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vmovaps %ymm1, %ymm8
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm2, %ymm0
+; AVX2-ONLY-NEXT: vmovaps %ymm3, %ymm7
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm2, %ymm4
+; AVX2-ONLY-NEXT: vmovaps %ymm1, %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm1
-; AVX2-ONLY-NEXT: vmovaps %ymm4, %ymm5
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm3, %ymm1
-; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm3
-; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 16(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm2, %ymm2
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 16(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm4 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm2 = [1,5,1,5,1,5,1,5]
-; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm13, %ymm2, %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 176(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [1,5,1,5]
-; AVX2-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm7, %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm2, %ymm0
+; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm2, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 176(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm2, %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm2, %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm2, %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 432(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm7, %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermps %ymm13, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vpermps %ymm12, %ymm2, %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 688(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm7, %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm9, %ymm2, %ymm4
+; AVX2-ONLY-NEXT: vmovaps %ymm9, %ymm10
+; AVX2-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 432(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm2, %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermps %ymm11, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm2, %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 944(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vmovaps %xmm6, (%rsp) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm7, %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm2, %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vpermps %ymm13, %ymm2, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 688(%rdi), %xmm5
; AVX2-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 48(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm7, %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm2, %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermps %ymm9, %ymm2, %ymm0
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm2, %ymm0
+; AVX2-ONLY-NEXT: vmovaps %ymm8, %ymm11
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm2, %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 304(%rdi), %xmm5
-; AVX2-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm7, %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm2, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 944(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vmovaps %xmm5, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm2, %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermps %ymm12, %ymm2, %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 560(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm7, %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm2, %ymm0
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm2, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 48(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm4 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm2, %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-ONLY-NEXT: vpermps %ymm6, %ymm2, %ymm0
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm2, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm7, %ymm1
-; AVX2-ONLY-NEXT: vmovaps 816(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps 304(%rdi), %xmm3
; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm2, %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermps %ymm13, %ymm2, %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermps %ymm12, %ymm2, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 560(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm2, %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm2, %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm2, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm2, %ymm2
+; AVX2-ONLY-NEXT: vmovaps 816(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm0 = [2,6,2,6,2,6,2,6]
-; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm3 = xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [2,6,2,6]
-; AVX2-ONLY-NEXT: # xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm0, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm0, %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm0, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm13, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermps %ymm13, %ymm0, %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm0, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermps %ymm11, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm0, %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm0, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermps %ymm9, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovaps %ymm9, %ymm7
-; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm0, %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpermps %ymm12, %ymm0, %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps %ymm6, %ymm8
; AVX2-ONLY-NEXT: vpermps %ymm6, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm0, %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm0, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm0 = [3,7,3,7,3,7,3,7]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm13, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vpermps %ymm12, %ymm0, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm0, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm3 = xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [3,7,3,7]
-; AVX2-ONLY-NEXT: # xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm11[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm4 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm0, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm4[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm0 = [3,7,3,7,3,7,3,7]
+; AVX2-ONLY-NEXT: vpermps %ymm14, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm0, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm14[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm0, %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm14[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm15, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vpermps %ymm10, %ymm0, %ymm14
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm14[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm14 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm14 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm14[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm11, %ymm0, %ymm11
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm0, %ymm14
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm14[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps (%rsp), %xmm5, %xmm14 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm14 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm14[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT: vpermps %ymm3, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm14[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpermps %ymm13, %ymm0, %ymm14
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm14[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm14 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm14 = xmm6[2],mem[2],xmm6[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm6, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm0, %ymm14
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm14 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm14 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm14[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm13, %ymm0, %ymm10
+; AVX2-ONLY-NEXT: vpermps %ymm12, %ymm0, %ymm14
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm14[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm13 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm13 = xmm5[2],mem[2],xmm5[3],mem[3]
; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps (%rsp), %xmm6, %xmm14 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm14 = xmm6[2],mem[2],xmm6[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm0, %ymm4
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm0, %ymm5
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm14[0,1],xmm13[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm13[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm14[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm0, %ymm5
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm14 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm14 = xmm6[2],mem[2],xmm6[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm14[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm14 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm14 = xmm6[2],mem[2],xmm6[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm14[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm0, %ymm8
; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
-; AVX2-ONLY-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 192(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 128(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 64(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, (%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 224(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 160(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 96(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 192(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 128(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 64(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, (%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 224(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 160(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 96(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 192(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 128(%rcx)
-; AVX2-ONLY-NEXT: vmovaps %ymm9, 64(%rcx)
-; AVX2-ONLY-NEXT: vmovaps %ymm10, (%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 224(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 160(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 96(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 192(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 128(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 224(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 160(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 192(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 128(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 224(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 160(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 192(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 128(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%rcx)
+; AVX2-ONLY-NEXT: vmovaps %ymm9, (%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 224(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 160(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%rcx)
; AVX2-ONLY-NEXT: vmovaps %ymm0, 192(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm12, 128(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm4, (%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm13, 224(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm10, 128(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 64(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm1, (%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm11, 224(%r8)
; AVX2-ONLY-NEXT: vmovaps %ymm2, 160(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 96(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm11, 32(%r8)
-; AVX2-ONLY-NEXT: addq $1992, %rsp # imm = 0x7C8
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 96(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
+; AVX2-ONLY-NEXT: addq $1944, %rsp # imm = 0x798
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
index 5d3f85d6ab6924..0c229457fb6024 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
@@ -569,82 +569,78 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-SLOW-NEXT: vmovaps 128(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovaps 160(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm3
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm4
-; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm7
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm2 = <0,6,4,u>
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm2, %ymm2
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm7[0,1],ymm5[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm10 = [4,2,4,2,4,2,4,2]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm10, %ymm10
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm10 = <1,7,5,u>
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm10, %ymm6
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm8 = [5,3,5,3,5,3,5,3]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm8, %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm7[0,1],ymm5[2,3],ymm7[4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
-; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm9
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm10 = xmm9[2,3,2,3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm11 = ymm4[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm11[1,2,3],ymm10[4],ymm11[5,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm1[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm11 = ymm0[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm5[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm7[1],ymm10[2,3,4],ymm7[5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm3 = <0,6,4,u>
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm4[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm6[0,1],ymm5[0,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm8 = [4,2,4,2,4,2,4,2]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm8, %ymm11
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm11 = <1,7,5,u>
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm11, %ymm7
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm9 = [5,3,5,3,5,3,5,3]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm9, %ymm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm9 = xmm9[3,3,3,3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1,2,3],ymm9[4],ymm11[5,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm1[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm11 = ymm0[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm10
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm11 = [4,2,4,2]
-; AVX2-SLOW-NEXT: # xmm11 = mem[0,0]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm11, %ymm4
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm11
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm12 = xmm11[2,3,2,3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1,2,3],ymm12[4],ymm13[5,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1,2],ymm10[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm12 = ymm1[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm0[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6],ymm12[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm12[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm12 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0],ymm6[1],ymm12[2,3,4],ymm6[5],ymm12[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,2,0,3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm11 = xmm11[3,3,3,3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0],ymm13[1,2,3],ymm11[4],ymm13[5,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm12 = ymm1[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm0[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6],ymm12[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm12[5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm12
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm5[0,1],ymm12[2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm6[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm1 = [0,2,0,6,0,2,0,6]
; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,0,1]
; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm1, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm5[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm10[3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm5 = [5,3,5,3]
-; AVX2-SLOW-NEXT: # xmm5 = mem[0,0]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm5, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm12[3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm9, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,3,1,7,0,3,1,7]
; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,0,1]
; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm4, %ymm0
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rsi)
-; AVX2-SLOW-NEXT: vmovaps %ymm6, (%rdx)
-; AVX2-SLOW-NEXT: vmovaps %ymm8, (%rcx)
-; AVX2-SLOW-NEXT: vmovaps %ymm9, (%r8)
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rsi)
+; AVX2-SLOW-NEXT: vmovaps %ymm7, (%rdx)
+; AVX2-SLOW-NEXT: vmovaps %ymm10, (%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm11, (%r8)
; AVX2-SLOW-NEXT: vmovaps %ymm1, (%r9)
; AVX2-SLOW-NEXT: vmovaps %ymm0, (%rax)
; AVX2-SLOW-NEXT: vzeroupper
@@ -656,82 +652,78 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-NEXT: vmovaps 128(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovaps 160(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovaps 96(%rdi), %ymm5
-; AVX2-FAST-NEXT: vmovaps (%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovaps (%rdi), %ymm2
; AVX2-FAST-NEXT: vmovaps 32(%rdi), %ymm4
-; AVX2-FAST-NEXT: vmovaps 64(%rdi), %ymm7
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm2 = <0,6,4,u>
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm7[0,1],ymm5[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm10 = [4,2,4,2,4,2,4,2]
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm10, %ymm10
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm10 = <1,7,5,u>
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm10, %ymm6
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm8 = [5,3,5,3,5,3,5,3]
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm8, %ymm8
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm7[0,1],ymm5[2,3],ymm7[4,5],ymm5[6,7]
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm9 = [2,4,2,4,2,4,2,4]
-; AVX2-FAST-NEXT: vpermps %ymm8, %ymm9, %ymm8
-; AVX2-FAST-NEXT: vmovaps (%rdi), %xmm9
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm10 = xmm9[2,3,2,3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm11 = ymm4[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm11[1,2,3],ymm10[4],ymm11[5,6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm1[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm11 = ymm0[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm5[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm7[1],ymm10[2,3,4],ymm7[5],ymm10[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm9 = xmm9[3,3,3,3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1,2,3],ymm9[4],ymm11[5,6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm1[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm11 = ymm0[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm10
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm11 = [4,2,4,2]
-; AVX2-FAST-NEXT: # xmm11 = mem[0,0]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm11, %ymm4
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 64(%rdi), %ymm6
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm3 = <0,6,4,u>
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm4[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm7, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm6[0,1],ymm5[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm8 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm8 = [4,2,4,2,4,2,4,2]
+; AVX2-FAST-NEXT: vpermps %ymm10, %ymm8, %ymm11
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm11 = <1,7,5,u>
+; AVX2-FAST-NEXT: vpermps %ymm7, %ymm11, %ymm7
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm9 = [5,3,5,3,5,3,5,3]
+; AVX2-FAST-NEXT: vpermps %ymm10, %ymm9, %ymm10
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm11 = [2,4,2,4,2,4,2,4]
+; AVX2-FAST-NEXT: vpermps %ymm10, %ymm11, %ymm10
+; AVX2-FAST-NEXT: vmovaps (%rdi), %xmm11
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm12 = xmm11[2,3,2,3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm13 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1,2,3],ymm12[4],ymm13[5,6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1,2],ymm10[3,4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm12 = ymm1[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm13 = ymm0[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6],ymm12[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm12[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm12 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0],ymm6[1],ymm12[2,3,4],ymm6[5],ymm12[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,2,0,3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm11 = xmm11[3,3,3,3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm13 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0],ymm13[1,2,3],ymm11[4],ymm13[5,6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm12 = ymm1[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm13 = ymm0[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6],ymm12[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm12[5,6,7]
+; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm12
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm5[0,1],ymm12[2,3],ymm5[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm6[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm8, %ymm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm1 = [0,2,0,6,0,2,0,6]
; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1]
; AVX2-FAST-NEXT: vpermps %ymm0, %ymm1, %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm5[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm10[3],ymm4[4,5,6,7]
-; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm5 = [5,3,5,3]
-; AVX2-FAST-NEXT: # xmm5 = mem[0,0]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm5, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm12[3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm9, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,3,1,7,0,3,1,7]
; AVX2-FAST-NEXT: # ymm4 = mem[0,1,0,1]
; AVX2-FAST-NEXT: vpermps %ymm0, %ymm4, %ymm0
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vmovaps %ymm2, (%rsi)
-; AVX2-FAST-NEXT: vmovaps %ymm6, (%rdx)
-; AVX2-FAST-NEXT: vmovaps %ymm8, (%rcx)
-; AVX2-FAST-NEXT: vmovaps %ymm9, (%r8)
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT: vmovaps %ymm3, (%rsi)
+; AVX2-FAST-NEXT: vmovaps %ymm7, (%rdx)
+; AVX2-FAST-NEXT: vmovaps %ymm10, (%rcx)
+; AVX2-FAST-NEXT: vmovaps %ymm11, (%r8)
; AVX2-FAST-NEXT: vmovaps %ymm1, (%r9)
; AVX2-FAST-NEXT: vmovaps %ymm0, (%rax)
; AVX2-FAST-NEXT: vzeroupper
@@ -743,82 +735,78 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm7
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm2 = <0,6,4,u>
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm2, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm7[0,1],ymm5[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm10 = [4,2,4,2,4,2,4,2]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm10, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm10 = <1,7,5,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm10, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm8 = [5,3,5,3,5,3,5,3]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm8, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm7[0,1],ymm5[2,3],ymm7[4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm9
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm10 = xmm9[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm11 = ymm4[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm11[1,2,3],ymm10[4],ymm11[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm1[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm11 = ymm0[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm5[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm7[1],ymm10[2,3,4],ymm7[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm3 = <0,6,4,u>
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm4[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm3, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm6[0,1],ymm5[0,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm8 = [4,2,4,2,4,2,4,2]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm8, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm11 = <1,7,5,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm11, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm9 = [5,3,5,3,5,3,5,3]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm9, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm9 = xmm9[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1,2,3],ymm9[4],ymm11[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm1[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm11 = ymm0[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm10
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm11 = [4,2,4,2]
-; AVX2-FAST-PERLANE-NEXT: # xmm11 = mem[0,0]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm11, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm11
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm12 = xmm11[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1,2,3],ymm12[4],ymm13[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1,2],ymm10[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm12 = ymm1[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm0[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm12[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm12 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0],ymm6[1],ymm12[2,3,4],ymm6[5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm11 = xmm11[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0],ymm13[1,2,3],ymm11[4],ymm13[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm12 = ymm1[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm0[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm12[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm5[0,1],ymm12[2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm6[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm1 = [0,2,0,6,0,2,0,6]
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm5[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm10[3],ymm4[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm5 = [5,3,5,3]
-; AVX2-FAST-PERLANE-NEXT: # xmm5 = mem[0,0]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm5, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm12[3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm9, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,3,1,7,0,3,1,7]
; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm4, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, (%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, (%r8)
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm7, (%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm10, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm11, (%r8)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%r9)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, (%rax)
; AVX2-FAST-PERLANE-NEXT: vzeroupper
@@ -1367,539 +1355,557 @@ define void @load_i32_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX2-SLOW-LABEL: load_i32_stride6_vf16:
; AVX2-SLOW: # %bb.0:
-; AVX2-SLOW-NEXT: subq $232, %rsp
-; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %ymm9
+; AVX2-SLOW-NEXT: subq $360, %rsp # imm = 0x168
+; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %ymm8
; AVX2-SLOW-NEXT: vmovaps 224(%rdi), %ymm11
-; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %ymm3
-; AVX2-SLOW-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 128(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 160(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %ymm7
-; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm12
-; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm10
+; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm3
+; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm13
+; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm6
; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm14 = <0,6,4,u>
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3],ymm12[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm14, %ymm1
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm10[0,1],ymm7[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm8[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm13[4,5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm14, %ymm3
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm6[0,1],ymm15[0,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm7[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm7[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm0 = [4,2,4,2,4,2,4,2]
-; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm0, %ymm15
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm11[4,5],ymm3[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm14, %ymm14
-; AVX2-SLOW-NEXT: vmovaps 256(%rdi), %ymm6
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm6[0,1],ymm9[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm15 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %ymm14
-; AVX2-SLOW-NEXT: vmovaps 352(%rdi), %ymm15
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm0, %ymm12
+; AVX2-SLOW-NEXT: vmovaps %ymm0, %ymm5
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm8[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm11[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm14, %ymm12
+; AVX2-SLOW-NEXT: vmovaps 256(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm7[0,1],ymm8[0,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm14 = ymm1[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %ymm12
+; AVX2-SLOW-NEXT: vmovaps 352(%rdi), %ymm14
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm5, %ymm11
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm2 = <1,7,5,u>
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm2, %ymm4
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm9[3,4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm5 = [5,3,5,3,5,3,5,3]
-; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm5, %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm5, %ymm1
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm5, %ymm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm5, %ymm0
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm3
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm3[2,3,2,3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm5
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm5[2,3,2,3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1],ymm7[2,3],ymm10[4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm15[2,3],ymm6[4,5],ymm15[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm8[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %xmm0
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm11[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm9[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm9[2,3],ymm6[4,5],ymm9[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5],ymm8[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm15[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm14[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm14[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm12[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm3[3,3,3,3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm5[3,3,3,3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm13[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm7[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm10[1],ymm2[2,3,4],ymm10[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm15[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm6[1],ymm2[2,3,4],ymm6[5],ymm2[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm13[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm4, %ymm5
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm9[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1],ymm1[2,3,4],ymm6[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm7[1],ymm1[2,3,4],ymm7[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm14[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vblendps $240, (%rsp), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm11 = ymm11[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 272(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm4 = [4,2,4,2]
-; AVX2-SLOW-NEXT: # xmm4 = mem[0,0]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm4, %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm15[0,1],ymm14[2,3],ymm15[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm6 = [0,2,0,6,0,2,0,6]
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm6, %ymm9
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm9 = ymm12[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm10
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm7[0,1],ymm10[2,3],ymm7[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm12 = ymm12[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm4, %ymm4
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm12[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm8[2,3],ymm13[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm6, %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm2 = [5,3,5,3]
-; AVX2-SLOW-NEXT: # xmm2 = mem[0,0]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm2, %ymm6
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm13[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm3
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm3[2,3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm6
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm6 = [0,3,1,7,0,3,1,7]
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm6, %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm10[3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm2, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm6, %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 32(%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, (%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 32(%rdx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, (%rdx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 32(%rcx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, (%rcx)
-; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%r8)
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm5[2,3],ymm10[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm11 = [0,2,0,6,0,2,0,6]
+; AVX2-SLOW-NEXT: # ymm11 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm11, %ymm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm9[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 272(%rdi), %xmm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm5[0,1],ymm13[2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm0, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm14[0,1],ymm12[2,3],ymm14[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm11, %ymm11
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm11[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm4, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm2 = [0,3,1,7,0,3,1,7]
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm2, %ymm3
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm4, %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm2, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rcx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%r8)
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm3, (%r8)
-; AVX2-SLOW-NEXT: vmovaps %ymm0, 32(%r9)
-; AVX2-SLOW-NEXT: vmovaps %ymm4, (%r9)
+; AVX2-SLOW-NEXT: vmovaps %ymm8, 32(%r9)
+; AVX2-SLOW-NEXT: vmovaps %ymm1, (%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-SLOW-NEXT: vmovaps %ymm1, 32(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rax)
-; AVX2-SLOW-NEXT: addq $232, %rsp
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm0, (%rax)
+; AVX2-SLOW-NEXT: addq $360, %rsp # imm = 0x168
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: load_i32_stride6_vf16:
; AVX2-FAST: # %bb.0:
-; AVX2-FAST-NEXT: subq $200, %rsp
-; AVX2-FAST-NEXT: vmovaps 288(%rdi), %ymm9
-; AVX2-FAST-NEXT: vmovaps 224(%rdi), %ymm11
-; AVX2-FAST-NEXT: vmovaps 192(%rdi), %ymm6
-; AVX2-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 128(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: subq $328, %rsp # imm = 0x148
+; AVX2-FAST-NEXT: vmovaps 288(%rdi), %ymm10
+; AVX2-FAST-NEXT: vmovaps 224(%rdi), %ymm14
+; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 192(%rdi), %ymm11
+; AVX2-FAST-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 128(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 160(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 96(%rdi), %ymm7
+; AVX2-FAST-NEXT: vmovaps 96(%rdi), %ymm15
; AVX2-FAST-NEXT: vmovaps (%rdi), %ymm3
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 32(%rdi), %ymm12
-; AVX2-FAST-NEXT: vmovaps 64(%rdi), %ymm8
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm1 = <0,6,4,u>
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm3[0,1,2,3],ymm12[4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm5, %ymm1, %ymm4
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm8[0,1],ymm7[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm13 = ymm10[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm0 = [4,2,4,2,4,2,4,2]
-; AVX2-FAST-NEXT: vpermps %ymm13, %ymm0, %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm11[4,5],ymm6[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm1, %ymm14
-; AVX2-FAST-NEXT: vmovaps 256(%rdi), %ymm6
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm6[0,1],ymm9[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm15 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 320(%rdi), %ymm14
-; AVX2-FAST-NEXT: vmovaps 352(%rdi), %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3],ymm14[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
-; AVX2-FAST-NEXT: vpermps %ymm5, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm10[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm5 = [5,3,5,3,5,3,5,3]
-; AVX2-FAST-NEXT: vpermps %ymm13, %ymm5, %ymm10
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-NEXT: vmovaps 32(%rdi), %ymm13
+; AVX2-FAST-NEXT: vmovaps 64(%rdi), %ymm6
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm0 = <0,6,4,u>
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm13[4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm0, %ymm7
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm6[0,1],ymm15[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm7[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm1 = [4,2,4,2,4,2,4,2]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm1, %ymm12
+; AVX2-FAST-NEXT: vmovaps %ymm1, %ymm5
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm11[0,1,2,3],ymm14[4,5],ymm11[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm9
+; AVX2-FAST-NEXT: vmovaps 256(%rdi), %ymm7
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm7[0,1],ymm10[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm12 = ymm1[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm12[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 320(%rdi), %ymm12
+; AVX2-FAST-NEXT: vmovaps 352(%rdi), %ymm14
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm5, %ymm11
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm11 = <1,7,5,u>
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm11, %ymm4
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm9 = [5,3,5,3,5,3,5,3]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm9, %ymm3
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm5, %ymm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm11, %ymm2
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm9, %ymm0
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps (%rdi), %xmm0
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm13[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm8[0,1],ymm7[2,3],ymm8[4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm15[2,3],ymm6[4,5],ymm15[6,7]
; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm3 = [2,0,6,4,2,0,6,7]
; AVX2-FAST-NEXT: vpermps %ymm2, %ymm3, %ymm2
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm10[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm13[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm8[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm5[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm9[2,3],ymm6[4,5],ymm9[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1],ymm10[2,3],ymm7[4,5],ymm10[6,7]
; AVX2-FAST-NEXT: vpermps %ymm1, %ymm3, %ymm1
; AVX2-FAST-NEXT: vmovaps 192(%rdi), %xmm2
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm2[2,3,2,3]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm11[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm15[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm14[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm14[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm12[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2,3,4],ymm8[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1],ymm1[2,3,4],ymm6[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm13[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vmovaps %ymm13, %ymm4
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm5[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm2[3,3,3,3]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vmovaps %ymm11, %ymm8
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm9[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1],ymm1[2,3,4],ymm6[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm10[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm7[1],ymm1[2,3,4],ymm7[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm14[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm11 = ymm11[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 272(%rdi), %xmm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm5 = [4,2,4,2]
-; AVX2-FAST-NEXT: # xmm5 = mem[0,0]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm5, %ymm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm15[0,1],ymm14[2,3],ymm15[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm9 = [0,2,0,6,0,2,0,6]
-; AVX2-FAST-NEXT: # ymm9 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm9, %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm8 = ymm12[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm12
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1],ymm12[2,3],ymm7[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm8, %ymm5, %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm10[0,1],ymm4[2,3],ymm10[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm14, %ymm9, %ymm9
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm2 = [5,3,5,3]
-; AVX2-FAST-NEXT: # xmm2 = mem[0,0]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm2, %ymm9
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm9 = [0,3,1,7,0,3,1,7]
-; AVX2-FAST-NEXT: # ymm9 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm9, %ymm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm12[3],ymm6[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm8, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm6[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm14, %ymm9, %ymm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm6, 32(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm6, (%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 32(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm6, (%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 32(%rcx)
-; AVX2-FAST-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, (%rcx)
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = ymm13[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm3
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm3[2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm6
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1],ymm5[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm11 = [0,2,0,6,0,2,0,6]
+; AVX2-FAST-NEXT: # ymm11 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpermps %ymm6, %ymm11, %ymm13
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = ymm8[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 272(%rdi), %xmm13
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm5[0,1],ymm13[2,3],ymm5[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm7, %ymm0, %ymm10
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm14[0,1],ymm12[2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm12, %ymm11, %ymm11
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm11[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm9, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm2 = [0,3,1,7,0,3,1,7]
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpermps %ymm6, %ymm2, %ymm3
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm7, %ymm9, %ymm5
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm12, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm3, (%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm3, (%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm3, (%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%r8)
-; AVX2-FAST-NEXT: vmovaps %ymm13, (%r8)
-; AVX2-FAST-NEXT: vmovaps %ymm0, 32(%r9)
-; AVX2-FAST-NEXT: vmovaps %ymm5, (%r9)
+; AVX2-FAST-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm3, (%r8)
+; AVX2-FAST-NEXT: vmovaps %ymm10, 32(%r9)
+; AVX2-FAST-NEXT: vmovaps %ymm1, (%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm2, (%rax)
-; AVX2-FAST-NEXT: addq $200, %rsp
+; AVX2-FAST-NEXT: vmovaps %ymm2, 32(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm0, (%rax)
+; AVX2-FAST-NEXT: addq $328, %rsp # imm = 0x148
; AVX2-FAST-NEXT: vzeroupper
; AVX2-FAST-NEXT: retq
;
; AVX2-FAST-PERLANE-LABEL: load_i32_stride6_vf16:
; AVX2-FAST-PERLANE: # %bb.0:
-; AVX2-FAST-PERLANE-NEXT: subq $232, %rsp
-; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %ymm9
+; AVX2-FAST-PERLANE-NEXT: subq $360, %rsp # imm = 0x168
+; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %ymm8
; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rdi), %ymm11
-; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %ymm7
-; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm13
+; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm6
; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm14 = <0,6,4,u>
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3],ymm12[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm14, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm10[0,1],ymm7[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm8[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm13[4,5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm14, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm6[0,1],ymm15[0,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm7[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm7[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm0 = [4,2,4,2,4,2,4,2]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm0, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm11[4,5],ymm3[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm14, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vmovaps 256(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm6[0,1],ymm9[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm15 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %ymm14
-; AVX2-FAST-PERLANE-NEXT: vmovaps 352(%rdi), %ymm15
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm8[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm11[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm14, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vmovaps 256(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm7[0,1],ymm8[0,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm14 = ymm1[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %ymm12
+; AVX2-FAST-PERLANE-NEXT: vmovaps 352(%rdi), %ymm14
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm5, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm2 = <1,7,5,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm2, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm5 = [5,3,5,3,5,3,5,3]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm5, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm5, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm5, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm2, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm5, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm3
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm3[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm5[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1],ymm7[2,3],ymm10[4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm15[2,3],ymm6[4,5],ymm15[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm8[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm11[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm9[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm9[2,3],ymm6[4,5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5],ymm8[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm15[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm14[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm14[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm12[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm3[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm5[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm13[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm7[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm10[1],ymm2[2,3,4],ymm10[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm15[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm6[1],ymm2[2,3,4],ymm6[5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm13[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, %ymm5
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm9[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1],ymm1[2,3,4],ymm6[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm7[1],ymm1[2,3,4],ymm7[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm14[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, (%rsp), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm11 = ymm11[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 272(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm4 = [4,2,4,2]
-; AVX2-FAST-PERLANE-NEXT: # xmm4 = mem[0,0]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm4, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm15[0,1],ymm14[2,3],ymm15[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm6 = [0,2,0,6,0,2,0,6]
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm6, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm12[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm10
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm7[0,1],ymm10[2,3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm12 = ymm12[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm4, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm12[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm8[2,3],ymm13[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm6, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm2 = [5,3,5,3]
-; AVX2-FAST-PERLANE-NEXT: # xmm2 = mem[0,0]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm2, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm13[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm3[2,3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm6
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm6 = [0,3,1,7,0,3,1,7]
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm6, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm10[3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm2, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm6, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 32(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 32(%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 32(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%r8)
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm5[2,3],ymm10[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm11 = [0,2,0,6,0,2,0,6]
+; AVX2-FAST-PERLANE-NEXT: # ymm11 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm11, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm9[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 272(%rdi), %xmm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm5[0,1],ymm13[2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm0, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm14[0,1],ymm12[2,3],ymm14[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm11, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm11[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm4, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm2 = [0,3,1,7,0,3,1,7]
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm2, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm4, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm2, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rsi)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 32(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 32(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rax)
-; AVX2-FAST-PERLANE-NEXT: addq $232, %rsp
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, (%rax)
+; AVX2-FAST-PERLANE-NEXT: addq $360, %rsp # imm = 0x168
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
;
@@ -3151,12 +3157,12 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX2-SLOW-LABEL: load_i32_stride6_vf32:
; AVX2-SLOW: # %bb.0:
-; AVX2-SLOW-NEXT: subq $1160, %rsp # imm = 0x488
-; AVX2-SLOW-NEXT: vmovaps 480(%rdi), %ymm6
+; AVX2-SLOW-NEXT: subq $1256, %rsp # imm = 0x4E8
+; AVX2-SLOW-NEXT: vmovaps 480(%rdi), %ymm9
; AVX2-SLOW-NEXT: vmovaps 448(%rdi), %ymm10
; AVX2-SLOW-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 416(%rdi), %ymm11
-; AVX2-SLOW-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 416(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 128(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 160(%rdi), %ymm3
@@ -3169,411 +3175,410 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm5
; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm12 = <0,6,4,u>
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm12, %ymm0
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm8 = <0,6,4,u>
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm8, %ymm0
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm4[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm4 = [4,2,4,2,4,2,4,2]
-; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm4, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm10[0,1],ymm6[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-SLOW-NEXT: vmovaps 384(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm5[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm2 = [4,2,4,2,4,2,4,2]
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm2, %ymm1
+; AVX2-SLOW-NEXT: vmovaps %ymm2, %ymm11
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm11[4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm12, %ymm1
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 512(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm10[0,1],ymm9[0,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-SLOW-NEXT: vmovaps 384(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 544(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm4, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3],ymm7[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm8, %ymm0
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 512(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovaps 544(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm11, %ymm1
+; AVX2-SLOW-NEXT: vmovaps %ymm11, %ymm12
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 256(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm0[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vmovaps 224(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovaps 256(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovaps 224(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm12, %ymm10
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm15 = ymm3[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm10[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 352(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm4, %ymm14
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 608(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm8, %ymm14
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm9 = ymm10[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm14[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 576(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm12, %ymm15
-; AVX2-SLOW-NEXT: vmovaps 672(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovaps 352(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm11, %ymm7
+; AVX2-SLOW-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 608(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 640(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm1[0,1],ymm0[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm12[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 704(%rdi), %ymm0
-; AVX2-SLOW-NEXT: vmovaps 736(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps %ymm0, %ymm15
+; AVX2-SLOW-NEXT: vmovaps 576(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm0[4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm8, %ymm8
+; AVX2-SLOW-NEXT: vmovaps 672(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm4, %ymm0
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovaps 640(%rdi), %ymm9
+; AVX2-SLOW-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm9[0,1],ymm0[0,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm11 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm11[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 704(%rdi), %ymm11
+; AVX2-SLOW-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 736(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm0, %ymm9
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm9 = [5,3,5,3,5,3,5,3]
-; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm9, %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm9, %ymm4
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm12, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm8 = <1,7,5,u>
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm8, %ymm13
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm13[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm12 = [5,3,5,3,5,3,5,3]
+; AVX2-SLOW-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm12, %ymm6
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm8, %ymm4
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm3[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm9, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm12, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm12[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm9, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm8, %ymm0
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm10[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm12, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm8, %ymm0
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm12, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm3
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm3[2,3,2,3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1],ymm11[2,3],ymm12[4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm8[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 384(%rdi), %xmm1
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm12[2,3],ymm9[4,5],ymm12[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5],ymm2[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm8[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 384(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps 576(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3],ymm5[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %xmm0
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm0[2,3,2,3]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm13[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1,2,3],ymm2[4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm13[0,1],ymm14[2,3],ymm13[4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm15[0,1],ymm14[2,3],ymm15[4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm10[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm3[3,3,3,3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm11[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm12[1],ymm3[2,3,4],ymm12[5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm12, %ymm11
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, (%rsp), %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3],ymm5[4,5,6],ymm3[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 576(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm3 = xmm2[2,3,2,3]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0,1],ymm2[2,3],mem[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm15[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %xmm15
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm4 = xmm15[2,3,2,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm2[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm3[0,1],ymm11[2,3],ymm3[4,5],ymm11[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm10[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2,3],ymm1[4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm12[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm9[1],ymm4[2,3,4],ymm9[5],ymm4[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm6[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm7[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm8[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm13, %ymm3
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm14[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1],ymm1[2,3,4],ymm13[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vmovaps %ymm13, %ymm14
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm10[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3],ymm4[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm15[3,3,3,3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm11[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovaps %ymm3, %ymm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3,4],ymm3[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4],ymm15[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm15, %ymm14
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm12[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermilps $255, (%rsp), %xmm0 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2,3,4],ymm8[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm5[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = ymm8[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = ymm14[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm12[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 464(%rdi), %xmm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm11 = [4,2,4,2]
-; AVX2-SLOW-NEXT: # xmm11 = mem[0,0]
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm11, %ymm1
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm13[0,1],ymm10[2,3],ymm13[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
-; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm0, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm8 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = ymm7[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 272(%rdi), %xmm1
-; AVX2-SLOW-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm11, %ymm2
-; AVX2-SLOW-NEXT: vmovaps %ymm11, %ymm13
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm12[2,3],ymm9[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm0, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm11 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm11 = ymm3[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3],ymm15[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 656(%rdi), %xmm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm12[2,3],ymm8[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm13, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm0, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm9 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm12, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm9 = [0,2,0,6,0,2,0,6]
+; AVX2-SLOW-NEXT: # ymm9 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm9, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm11[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm11 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm6
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm12, %ymm1
+; AVX2-SLOW-NEXT: vmovaps %ymm12, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm1
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm9[0,1],ymm1[2,3],ymm9[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm10[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm13, %ymm4
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm10[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm10 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm10 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm9, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm4 = [5,3,5,3]
-; AVX2-SLOW-NEXT: # xmm4 = mem[0,0]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm13 = [0,3,1,7,0,3,1,7]
-; AVX2-SLOW-NEXT: # ymm13 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm13, %ymm14
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps $8, (%rsp), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm10 = ymm14[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 272(%rdi), %xmm3
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1],ymm3[2,3],ymm10[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm12, %ymm14
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm9, %ymm12
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3,4],ymm12[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm12 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm12 = ymm8[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 656(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm12[0,1],ymm0[2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm15, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm9, %ymm9
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm14[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm14, %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm7 = [0,3,1,7,0,3,1,7]
+; AVX2-SLOW-NEXT: # ymm7 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm7, %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = ymm6[0,1,2],mem[3],ymm6[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm11 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm11[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm7, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm10[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2],ymm3[3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm14, %ymm5
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm13, %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm12[3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm4, %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm13, %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm9[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2],ymm1[3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm4, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm13, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 96(%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 64(%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 96(%rdx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rdx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 64(%rdx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rdx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rcx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 96(%rcx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 64(%rcx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rcx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 96(%r8)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%r8)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 64(%r8)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, (%r8)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 96(%r9)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%r9)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, (%r9)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 64(%r9)
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm7, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm14, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm7, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 96(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 32(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 64(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, (%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 96(%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 32(%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 64(%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, (%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 96(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 64(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, (%rcx)
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 96(%r8)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 32(%r8)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 64(%r8)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, (%r8)
+; AVX2-SLOW-NEXT: vmovaps %ymm9, 96(%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 32(%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, (%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 64(%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 96(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm0, 64(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm1, (%rax)
-; AVX2-SLOW-NEXT: addq $1160, %rsp # imm = 0x488
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 96(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm6, 64(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%rax)
+; AVX2-SLOW-NEXT: addq $1256, %rsp # imm = 0x4E8
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: load_i32_stride6_vf32:
; AVX2-FAST: # %bb.0:
-; AVX2-FAST-NEXT: subq $1160, %rsp # imm = 0x488
-; AVX2-FAST-NEXT: vmovaps 480(%rdi), %ymm6
+; AVX2-FAST-NEXT: subq $1256, %rsp # imm = 0x4E8
+; AVX2-FAST-NEXT: vmovaps 480(%rdi), %ymm9
; AVX2-FAST-NEXT: vmovaps 448(%rdi), %ymm10
; AVX2-FAST-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 416(%rdi), %ymm11
-; AVX2-FAST-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 416(%rdi), %ymm7
+; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 128(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 160(%rdi), %ymm3
@@ -3586,154 +3591,156 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 64(%rdi), %ymm5
; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm12 = <0,6,4,u>
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm12, %ymm0
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm8 = <0,6,4,u>
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm13, %ymm8, %ymm0
; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm4[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm4 = [4,2,4,2,4,2,4,2]
-; AVX2-FAST-NEXT: vpermps %ymm8, %ymm4, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm10[0,1],ymm6[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm12 = [4,2,4,2,4,2,4,2]
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm12, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm10[0,1],ymm9[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FAST-NEXT: vmovaps 384(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm11[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm5, %ymm12, %ymm1
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 512(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm7[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm8, %ymm0
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 512(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 544(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm4, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 288(%rdi), %ymm0
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm12, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 256(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps 288(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm0[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vmovaps 224(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovaps 256(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 192(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm12, %ymm10
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm15 = ymm3[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm10[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 320(%rdi), %ymm1
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovaps 224(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 352(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovaps 192(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm10, %ymm4, %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm8, %ymm14
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm9 = ymm10[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm14[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 320(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 352(%rdi), %ymm7
+; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm14, %ymm12, %ymm7
+; AVX2-FAST-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 608(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 576(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm14, %ymm12, %ymm15
+; AVX2-FAST-NEXT: vmovaps 576(%rdi), %ymm7
+; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm0[4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm7, %ymm8, %ymm8
; AVX2-FAST-NEXT: vmovaps 672(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 640(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm1[0,1],ymm0[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm13 = ymm12[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 704(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 736(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm0
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vmovaps 640(%rdi), %ymm9
+; AVX2-FAST-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm9[0,1],ymm0[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm11 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm11[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 704(%rdi), %ymm11
+; AVX2-FAST-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 736(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm0, %ymm9
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm9 = [5,3,5,3,5,3,5,3]
-; AVX2-FAST-NEXT: vpermps %ymm8, %ymm9, %ymm8
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm5, %ymm0, %ymm5
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm11, %ymm12, %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm8 = <1,7,5,u>
+; AVX2-FAST-NEXT: vpermps %ymm13, %ymm8, %ymm13
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm9, %ymm4
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm3[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm10, %ymm9, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm13[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm13 = [5,3,5,3,5,3,5,3]
+; AVX2-FAST-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm13, %ymm5
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm8, %ymm3
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm4[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm13, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm14, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm12[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm9, %ymm1
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm8, %ymm0
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm10[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm14, %ymm13, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm7, %ymm8, %ymm0
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm11, %ymm13, %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps (%rdi), %xmm0
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm12[2,3],ymm9[4,5],ymm12[6,7]
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm3 = [2,0,6,4,2,0,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm3, %ymm2
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm9[0,1],ymm11[2,3],ymm9[4,5],ymm11[6,7]
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm10 = [2,0,6,4,2,0,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm10, %ymm3
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm2[0,0,0,0,4,4,4,4]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm8[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 384(%rdi), %xmm1
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm1[2,3,2,3]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1,2,3],ymm2[4],ymm4[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm13[0,1],ymm14[2,3],ymm13[4,5],ymm14[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm3, %ymm4
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm10, %ymm4
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm4 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, (%rsp), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 576(%rdi), %xmm2
-; AVX2-FAST-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm4 = xmm2[2,3,2,3]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 576(%rdi), %xmm3
+; AVX2-FAST-NEXT: vmovaps %xmm3, (%rsp) # 16-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm4 = xmm3[2,3,2,3]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm13[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0,1],ymm2[2,3],mem[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm5, %ymm3, %ymm5
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = ymm3[0,1],mem[2,3],ymm3[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm10, %ymm5
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
@@ -3742,36 +3749,37 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm4 = mem[0,1],ymm11[2,3],mem[4,5],ymm11[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm3, %ymm4
-; AVX2-FAST-NEXT: vmovaps 192(%rdi), %xmm3
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm5 = xmm3[2,3,2,3]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm2[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm10, %ymm4
+; AVX2-FAST-NEXT: vmovaps 192(%rdi), %xmm12
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm5 = xmm12[2,3,2,3]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm14[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4],ymm6[5,6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm15[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm10[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm3[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm7[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1,2,3],ymm0[4],ymm4[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm11[3,3,3,3,7,7,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm9[1],ymm4[2,3,4],ymm9[5],ymm4[6,7]
+; AVX2-FAST-NEXT: vmovaps %ymm9, %ymm11
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm7[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm2[0,1,0,1,4,5,4,5]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm8[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
@@ -3781,213 +3789,206 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm14[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1],ymm1[2,3,4],ymm13[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps %ymm13, %ymm14
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm6[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vmovups (%rsp), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm7[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3],ymm4[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm3[3,3,3,3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm12[3,3,3,3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm11[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm7[1],ymm1[2,3,4],ymm7[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm14[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4],ymm15[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vmovaps %ymm15, %ymm12
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm15[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm3[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: vpermilps $255, (%rsp), %xmm0 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1],ymm1[2,3,4],ymm13[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm5[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm2 = ymm8[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm14[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 464(%rdi), %xmm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm11 = [4,2,4,2]
-; AVX2-FAST-NEXT: # xmm11 = mem[0,0]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm11, %ymm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm13[0,1],ymm10[2,3],ymm13[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
-; AVX2-FAST-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm14, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm8 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm7[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 272(%rdi), %xmm1
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm8, %ymm11, %ymm2
-; AVX2-FAST-NEXT: vmovaps %ymm11, %ymm13
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm12[2,3],ymm9[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm11 = ymm3[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3],ymm15[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 656(%rdi), %xmm12
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm12[2,3],ymm8[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm13, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm9 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm12, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm9 = [0,2,0,6,0,2,0,6]
+; AVX2-FAST-NEXT: # ymm9 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm9, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = ymm11[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm11 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm6
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm11, %ymm12, %ymm1
+; AVX2-FAST-NEXT: vmovaps %ymm12, %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm9[0,1],ymm1[2,3],ymm9[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm10[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm13, %ymm4
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm10[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm10 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm10, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm9, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm4 = [5,3,5,3]
-; AVX2-FAST-NEXT: # xmm4 = mem[0,0]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm13 = [0,3,1,7,0,3,1,7]
-; AVX2-FAST-NEXT: # ymm13 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm14, %ymm13, %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 272(%rdi), %xmm3
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1],ymm3[2,3],ymm10[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm12, %ymm14
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm9, %ymm12
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3,4],ymm12[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm12 = ymm13[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 656(%rdi), %xmm0
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm12[0,1],ymm0[2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm15, %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = ymm13[0,1],mem[2,3],ymm13[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm15, %ymm9, %ymm9
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm14[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermps %ymm11, %ymm14, %ymm7
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm7 = [0,3,1,7,0,3,1,7]
+; AVX2-FAST-NEXT: # ymm7 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm7, %ymm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = ymm6[0,1,2],mem[3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm11[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm7, %ymm8
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm8 = ymm10[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2],ymm3[3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm14, %ymm5
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm13, %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm12[3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm4, %ymm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm13, %ymm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm9[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2],ymm1[3],ymm6[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm4, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm10, %ymm13, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 96(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 32(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 64(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, (%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 96(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 32(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 64(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, (%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 32(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 96(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 64(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, (%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 96(%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 32(%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 64(%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, (%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 96(%r9)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 32(%r9)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, (%r9)
-; AVX2-FAST-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm2, 64(%r9)
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm7, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm14, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm15, %ymm7, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 96(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, (%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 96(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, (%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 96(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, (%rcx)
+; AVX2-FAST-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 96(%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, (%r8)
+; AVX2-FAST-NEXT: vmovaps %ymm9, 96(%r9)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%r9)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, (%r9)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm0, 64(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm1, (%rax)
-; AVX2-FAST-NEXT: addq $1160, %rsp # imm = 0x488
+; AVX2-FAST-NEXT: vmovaps %ymm0, 96(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm2, 32(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm6, 64(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm4, (%rax)
+; AVX2-FAST-NEXT: addq $1256, %rsp # imm = 0x4E8
; AVX2-FAST-NEXT: vzeroupper
; AVX2-FAST-NEXT: retq
;
; AVX2-FAST-PERLANE-LABEL: load_i32_stride6_vf32:
; AVX2-FAST-PERLANE: # %bb.0:
-; AVX2-FAST-PERLANE-NEXT: subq $1160, %rsp # imm = 0x488
-; AVX2-FAST-PERLANE-NEXT: vmovaps 480(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT: subq $1256, %rsp # imm = 0x4E8
+; AVX2-FAST-PERLANE-NEXT: vmovaps 480(%rdi), %ymm9
; AVX2-FAST-PERLANE-NEXT: vmovaps 448(%rdi), %ymm10
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 416(%rdi), %ymm11
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 416(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%rdi), %ymm3
@@ -4000,400 +4001,399 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm5
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm12 = <0,6,4,u>
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm12, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm8 = <0,6,4,u>
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm8, %ymm0
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm4[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm4 = [4,2,4,2,4,2,4,2]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm4, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm10[0,1],ymm6[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 384(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm5[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm2 = [4,2,4,2,4,2,4,2]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm2, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm11[4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm12, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 512(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm10[0,1],ymm9[0,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 384(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 544(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm4, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3],ymm7[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm8, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 512(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps 544(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm11, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm11, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 256(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm0[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps 256(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm12, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm15 = ymm3[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm10[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm8, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm9 = ymm10[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm14[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 352(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm4, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 352(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm11, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 608(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 576(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm12, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vmovaps 576(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm0[4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm8, %ymm8
; AVX2-FAST-PERLANE-NEXT: vmovaps 672(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 640(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm1[0,1],ymm0[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm12[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 704(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovaps 736(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm4, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 640(%rdi), %ymm9
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm9[0,1],ymm0[0,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm11 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm11[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 704(%rdi), %ymm11
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 736(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm0, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm9 = [5,3,5,3,5,3,5,3]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm9, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm9, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm12, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm8 = <1,7,5,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm8, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm13[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm12 = [5,3,5,3,5,3,5,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm12, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm8, %ymm4
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm3[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm9, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm12, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm12[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm9, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm8, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm10[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm12, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm8, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm12, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm3[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1],ymm11[2,3],ymm12[4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm8[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 384(%rdi), %xmm1
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm12[2,3],ymm9[4,5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm8[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 576(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3],ymm5[4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 384(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm2 = xmm0[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm13[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1,2,3],ymm2[4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm13[0,1],ymm14[2,3],ymm13[4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm15[0,1],ymm14[2,3],ymm15[4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm10[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm2 = xmm3[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm11[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm12[1],ymm3[2,3,4],ymm12[5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm12, %ymm11
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, (%rsp), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3],ymm5[4,5,6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 576(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm3 = xmm2[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1],ymm2[2,3],mem[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm15[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %xmm15
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm4 = xmm15[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm2[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm3[0,1],ymm11[2,3],ymm3[4,5],ymm11[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm10[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2,3],ymm1[4],ymm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm12[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm9[1],ymm4[2,3,4],ymm9[5],ymm4[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm6[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm7[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm8[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm13, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm14[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1],ymm1[2,3,4],ymm13[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm13, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm10[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3],ymm4[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm15[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm11[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3,4],ymm3[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4],ymm15[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm15, %ymm14
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm12[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, (%rsp), %xmm0 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2,3,4],ymm8[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm5[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm8[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm14[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm12[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 464(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm11 = [4,2,4,2]
-; AVX2-FAST-PERLANE-NEXT: # xmm11 = mem[0,0]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm11, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm13[0,1],ymm10[2,3],ymm13[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm8 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm7[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 272(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm11, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm11, %ymm13
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm12[2,3],ymm9[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm11 = ymm3[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3],ymm15[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 656(%rdi), %xmm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm12[2,3],ymm8[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm13, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm12, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm9 = [0,2,0,6,0,2,0,6]
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm9, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm11[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm11 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm6
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm12, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm12, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm9[0,1],ymm1[2,3],ymm9[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm10[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm13, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm10[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm10 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm9, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm4 = [5,3,5,3]
-; AVX2-FAST-PERLANE-NEXT: # xmm4 = mem[0,0]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm13 = [0,3,1,7,0,3,1,7]
-; AVX2-FAST-PERLANE-NEXT: # ymm13 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm13, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps $8, (%rsp), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm10 = ymm14[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 272(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1],ymm3[2,3],ymm10[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm12, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm9, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3,4],ymm12[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm12 = ymm8[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 656(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm12[0,1],ymm0[2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm15, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm9, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm14[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm14, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm7 = [0,3,1,7,0,3,1,7]
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm7, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm6[0,1,2],mem[3],ymm6[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm11[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm7, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm10[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2],ymm3[3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm14, %ymm5
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm13, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm12[3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm4, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm13, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm9[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2],ymm1[3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm4, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm13, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 96(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 64(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 96(%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 64(%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 96(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 64(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 96(%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 64(%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 96(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 64(%r9)
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm7, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm14, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm7, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 96(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 96(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 96(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 96(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 96(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 96(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 64(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%rax)
-; AVX2-FAST-PERLANE-NEXT: addq $1160, %rsp # imm = 0x488
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 96(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 64(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%rax)
+; AVX2-FAST-PERLANE-NEXT: addq $1256, %rsp # imm = 0x4E8
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
;
@@ -6945,45 +6945,47 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX2-SLOW-LABEL: load_i32_stride6_vf64:
; AVX2-SLOW: # %bb.0:
-; AVX2-SLOW-NEXT: subq $2504, %rsp # imm = 0x9C8
-; AVX2-SLOW-NEXT: vmovaps 672(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovaps 640(%rdi), %ymm3
+; AVX2-SLOW-NEXT: subq $2536, %rsp # imm = 0x9E8
+; AVX2-SLOW-NEXT: vmovaps 672(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vmovaps 640(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 608(%rdi), %ymm3
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 608(%rdi), %ymm4
-; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vmovups %ymm5, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 352(%rdi), %ymm6
; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %ymm7
; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 256(%rdi), %ymm9
-; AVX2-SLOW-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups %ymm9, (%rsp) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 224(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm8 = <0,6,4,u>
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm8, %ymm0
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm8, %ymm0
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[0,1],ymm7[0,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm7[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm5 = [4,2,4,2,4,2,4,2]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm5, %ymm1
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm2 = [4,2,4,2,4,2,4,2]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT: vmovaps %ymm2, %ymm6
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm3[0,1],ymm2[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[0,1],ymm4[0,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm4[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 576(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm8, %ymm0
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
@@ -6994,7 +6996,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm5, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm6, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1056(%rdi), %ymm1
@@ -7019,7 +7021,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm5, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm6, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1440(%rdi), %ymm1
@@ -7044,7 +7046,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm5, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm6, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %ymm1
@@ -7052,21 +7054,21 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm8, %ymm0
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm15[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm8, %ymm0
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,2,2,2,4,6,6,6]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 128(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 160(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm5, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm6, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 480(%rdi), %ymm1
@@ -7074,21 +7076,22 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovaps 448(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovaps 416(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 384(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm8, %ymm0
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm8, %ymm0
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,2,2,2,4,6,6,6]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 512(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 544(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm5, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm6, %ymm1
+; AVX2-SLOW-NEXT: vmovaps %ymm6, %ymm12
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 864(%rdi), %ymm1
@@ -7096,114 +7099,117 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovaps 832(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovaps 800(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 768(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm8, %ymm0
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm0
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2,2,2,4,6,6,6]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 896(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 928(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm5, %ymm1
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovaps 928(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm12, %ymm6
+; AVX2-SLOW-NEXT: vmovaps %ymm12, %ymm14
+; AVX2-SLOW-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm6[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1184(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 1152(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm0
-; AVX2-SLOW-NEXT: vmovaps 1248(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vmovaps 1152(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm8, %ymm0
+; AVX2-SLOW-NEXT: vmovaps 1248(%rdi), %ymm9
+; AVX2-SLOW-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 1216(%rdi), %ymm8
; AVX2-SLOW-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 1216(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm8[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 1280(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 1312(%rdi), %ymm13
-; AVX2-SLOW-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm5, %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm8[0,1],ymm9[0,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1280(%rdi), %ymm9
+; AVX2-SLOW-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 1312(%rdi), %ymm12
+; AVX2-SLOW-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm12[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm14, %ymm9
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm9[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm5 = <1,7,5,u>
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm5, %ymm0
-; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm13 = mem[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm0 = [5,3,5,3,5,3,5,3]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm9[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm9 = [5,3,5,3,5,3,5,3]
+; AVX2-SLOW-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm12 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm14[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm5, %ymm12
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm15[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm0, %ymm11
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm5, %ymm9
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm10[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm0, %ymm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm5, %ymm4
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm5, %ymm2
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm8[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm12 = ymm13[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm9, %ymm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm9, %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm3[1,3,2,3,5,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm9, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm8[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm9, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %xmm4
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm4[2,3,2,3]
+; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %xmm5
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm5[2,3,2,3]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm1[0,1],mem[2,3],ymm1[4,5],mem[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, (%rsp), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
@@ -7216,8 +7222,8 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm1[0,1],mem[2,3],ymm1[4,5],mem[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
@@ -7232,8 +7238,8 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 960(%rdi), %xmm2
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
@@ -7244,52 +7250,52 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2],ymm1[3],ymm5[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3],ymm4[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1344(%rdi), %xmm1
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1,2,3],ymm0[4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm15[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1,2,3],ymm0[4],ymm4[5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm13[2,3],ymm5[4,5],ymm13[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
+; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,1],ymm13[2,3],mem[4,5],ymm13[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,0,0,0,4,4,4,4]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm14[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm15[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm14[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3],ymm6[4,5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm0
; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1,2,3],ymm0[4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,1],ymm5[2,3],mem[4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1,2,3],ymm0[4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,0,0,0,4,4,4,4]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3],ymm6[4,5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 384(%rdi), %xmm0
; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7297,99 +7303,99 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm6[1,2,3],ymm0[4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,1],ymm5[2,3],mem[4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = ymm4[0,1],mem[2,3],ymm4[4,5],mem[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm6[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm6 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6],ymm6[7]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3],ymm8[4,5,6],ymm6[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 768(%rdi), %xmm0
; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm0[2,3,2,3]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4],ymm7[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0,0,0,0,4,4,4,4]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3],ymm8[4,5,6],ymm7[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 1152(%rdi), %xmm12
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm7 = xmm12[2,3,2,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm9[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1,2,3],ymm7[4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm8[1,2,3],ymm6[4],ymm8[5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm11[0,1],ymm0[2,3],ymm11[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm6[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm5[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm10 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3],ymm10[4,5,6],ymm8[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm8[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm4 = xmm4[3,3,3,3]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm7[1,2,3],ymm4[4],ymm7[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm3[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps $244, (%rsp), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 1152(%rdi), %xmm12
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm8 = xmm12[2,3,2,3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm6[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm10[1,2,3],ymm8[4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm9[0,1],ymm0[2,3],ymm9[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm10[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm4[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm11 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,3,3,3]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3],ymm8[4,5,6],ymm7[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm8[1,2,3],ymm5[4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps $34, (%rsp), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm10 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3],ymm10[4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm8[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm3 = mem[3,3,3,3]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm5[1,2,3],ymm3[4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3],ymm7[4,5,6],ymm4[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm8[0,1,2],ymm5[3],ymm8[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
@@ -7404,15 +7410,14 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3],ymm5[4,5,6],ymm3[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm13[3,3,3,3,7,7,7,7]
; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
@@ -7420,302 +7425,298 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm14[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm14[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm12[3,3,3,3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm11[1],ymm2[2,3,4],ymm11[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm9[1],ymm2[2,3,4],ymm9[5],ymm2[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm5[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm13[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm12[1],ymm1[2,3,4],ymm12[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1],ymm1[2,3,4],ymm6[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm7[1],ymm1[2,3,4],ymm7[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,1,0,1,4,5,4,5]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4],ymm15[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm5[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm3[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 272(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm6[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm1 = [4,2,4,2]
-; AVX2-SLOW-NEXT: # xmm1 = mem[0,0]
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm1, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm5 = [0,2,0,6,0,2,0,6]
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm5, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm4[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm14, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 464(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm8[2,3],ymm9[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm5, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
+; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 656(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vmovaps 272(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm14, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm5, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm8[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = ymm10[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm11[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 848(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vmovaps 464(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm14, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1],ymm9[2,3],ymm10[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm12[2,3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm5, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 1040(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm13 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm5, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 656(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm14, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 1232(%rdi), %xmm11
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1],ymm11[2,3],ymm12[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm9 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm5, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 1424(%rdi), %xmm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm5, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 848(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm14, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm10 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm10 = ymm15[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm13 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm12 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1040(%rdi), %xmm11
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1],ymm11[2,3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm14, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1],ymm3[2,3],ymm10[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm1, %ymm1
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm10 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1232(%rdi), %xmm6
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm6[2,3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm14, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm5, %ymm0
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1424(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm14, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm0 = mem[1,1,1,1,5,5,5,5]
; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm1 = [5,3,5,3]
-; AVX2-SLOW-NEXT: # xmm1 = mem[0,0]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm0[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps (%rsp), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps $8, (%rsp), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm0, %ymm13
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm12 = ymm12[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3],ymm12[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm13[0,1,2],ymm11[3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm9, %ymm12
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm0, %ymm9
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm1, %ymm1
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm0, %ymm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm8[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm9, %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm9, %ymm3
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm1, 192(%rsi)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -7797,788 +7798,785 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm1, (%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-SLOW-NEXT: vmovaps %ymm6, 224(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm9, 192(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm13, 160(%rax)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm1, 128(%rax)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm1, 96(%rax)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm1, 64(%rax)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm1, 32(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 224(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm5, 192(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm10, 160(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm14, 128(%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 96(%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 64(%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, (%rax)
-; AVX2-SLOW-NEXT: addq $2504, %rsp # imm = 0x9C8
+; AVX2-SLOW-NEXT: addq $2536, %rsp # imm = 0x9E8
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: load_i32_stride6_vf64:
; AVX2-FAST: # %bb.0:
; AVX2-FAST-NEXT: subq $2504, %rsp # imm = 0x9C8
-; AVX2-FAST-NEXT: vmovaps 672(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovaps 640(%rdi), %ymm3
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 608(%rdi), %ymm4
-; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 320(%rdi), %ymm5
+; AVX2-FAST-NEXT: vmovaps 672(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovaps 640(%rdi), %ymm5
; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 352(%rdi), %ymm6
+; AVX2-FAST-NEXT: vmovaps 608(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 320(%rdi), %ymm6
; AVX2-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 288(%rdi), %ymm7
+; AVX2-FAST-NEXT: vmovaps 352(%rdi), %ymm7
; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 288(%rdi), %ymm8
+; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 256(%rdi), %ymm9
-; AVX2-FAST-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 224(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 192(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovups %ymm9, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 224(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 192(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm0 = <0,6,4,u>
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm14, %ymm0, %ymm1
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm9[0,1],ymm8[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm2[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm15[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm13 = [4,2,4,2,4,2,4,2]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm13, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm8 = <0,6,4,u>
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm8, %ymm0
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[0,1],ymm7[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm4[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 576(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm5 = [4,2,4,2,4,2,4,2]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm5, %ymm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm0, %ymm1
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 704(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm3[0,1],ymm2[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vmovaps 736(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 576(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm0
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 704(%rdi), %ymm1
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm13, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 736(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 1056(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1024(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm5, %ymm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1056(%rdi), %ymm1
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm2[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 992(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1024(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovaps 960(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 992(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 960(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm0
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 1088(%rdi), %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1120(%rdi), %ymm2
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm0, %ymm1
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm3[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1088(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm5, %ymm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1440(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps 1120(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm13, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1408(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovaps 1440(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1376(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1344(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps 1408(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm0
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 1472(%rdi), %ymm1
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm2[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1376(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1504(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 1344(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm5, %ymm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 96(%rdi), %ymm1
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm0, %ymm1
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm3[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1472(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1504(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm13, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 64(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps (%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 32(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps 96(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 64(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm12, %ymm8, %ymm0
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 128(%rdi), %ymm1
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm2[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps (%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 160(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 32(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm5, %ymm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 480(%rdi), %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm12, %ymm0, %ymm1
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm3[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 128(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 160(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm11, %ymm13, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 448(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps 416(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 384(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps 480(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 448(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm8, %ymm0
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 512(%rdi), %ymm1
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm2[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vmovaps 416(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 544(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 384(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm5, %ymm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 864(%rdi), %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm0, %ymm1
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm10[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 512(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 544(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm13, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 832(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps 800(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 768(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps 864(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 832(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm8, %ymm0
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 896(%rdi), %ymm1
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm2[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vmovaps 800(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 928(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 768(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm5, %ymm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1184(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1152(%rdi), %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm0, %ymm1
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 896(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 928(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm13, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm8, %ymm0
-; AVX2-FAST-NEXT: vmovaps 1248(%rdi), %ymm8
-; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1184(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1152(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vmovaps 1248(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 1216(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm8[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm2[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 1280(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1312(%rdi), %ymm13
-; AVX2-FAST-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm5, %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-FAST-NEXT: vmovaps 1312(%rdi), %ymm7
+; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm13, %ymm7
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm5 = <1,7,5,u>
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm5, %ymm0
-; AVX2-FAST-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm0 = [5,3,5,3,5,3,5,3]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
+; AVX2-FAST-NEXT: vpermps %ymm14, %ymm0, %ymm7
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm14 = ymm15[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm7 = [5,3,5,3,5,3,5,3]
+; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm12, %ymm5, %ymm12
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm13 = ymm14[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm0, %ymm11
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm12, %ymm0, %ymm12
+; AVX2-FAST-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm11, %ymm7, %ymm11
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
; AVX2-FAST-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm5, %ymm9
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm0, %ymm9
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm10[1,3,2,3,5,7,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm0, %ymm7
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm5, %ymm4
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm7, %ymm8
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm0, %ymm5
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm5, %ymm2
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm8[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 192(%rdi), %xmm2
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm6 = [2,0,6,4,2,0,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm6, %ymm1
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm7, %ymm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm7, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 576(%rdi), %xmm1
+; AVX2-FAST-NEXT: vmovaps 192(%rdi), %xmm1
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1,2,3],ymm0[4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm13[0,1],ymm12[2,3],ymm13[4,5],ymm12[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm6, %ymm3
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm0 = [2,0,6,4,2,0,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm3
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm3 = mem[0,0,0,0,4,4,4,4]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 960(%rdi), %xmm0
-; AVX2-FAST-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 576(%rdi), %xmm3
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm3[2,3,2,3]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1,2,3],ymm0[4],ymm4[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm4 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm6, %ymm4
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1,2,3],ymm2[4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = ymm4[0,1],mem[2,3],ymm4[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm0, %ymm4
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm4 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm15[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1344(%rdi), %xmm0
-; AVX2-FAST-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX2-FAST-NEXT: vpermilps $224, (%rsp), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1,2,3],ymm0[4],ymm5[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm4[0,1],ymm14[2,3],ymm4[4,5],ymm14[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm5, %ymm6, %ymm5
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm4[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2],ymm5[3],ymm7[4,5,6],ymm5[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps (%rdi), %xmm0
-; AVX2-FAST-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1,2,3],ymm0[4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 960(%rdi), %xmm2
+; AVX2-FAST-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm5[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1,2,3],ymm2[4],ymm4[5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm5, %ymm6, %ymm5
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2],ymm5[3],ymm7[4,5,6],ymm5[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 384(%rdi), %xmm0
-; AVX2-FAST-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm7[1,2,3],ymm0[4],ymm7[5,6,7]
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm0, %ymm4
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3],ymm6[4,5,6],ymm4[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1344(%rdi), %xmm2
+; AVX2-FAST-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm2[2,3,2,3]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm6, %ymm7
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm6[1,2,3],ymm2[4],ymm6[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm15[0,1],ymm14[2,3],ymm15[4,5],ymm14[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm6
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3],ymm8[4,5,6],ymm6[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps (%rdi), %xmm2
+; AVX2-FAST-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm6[1,2,3],ymm2[4],ymm6[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1],ymm6[2,3],mem[4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm6
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,0,0,0,4,4,4,4]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3],ymm9[4,5,6],ymm7[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 768(%rdi), %xmm0
-; AVX2-FAST-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm7 = xmm0[2,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0,1,2],ymm6[3],ymm9[4,5,6],ymm6[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 384(%rdi), %xmm2
+; AVX2-FAST-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm2[2,3,2,3]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0],ymm9[1,2,3],ymm7[4],ymm9[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm9 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm6, %ymm9
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm9[1,2,3],ymm2[4],ymm9[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[0,1],ymm6[2,3],mem[4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm0, %ymm9
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm9 = mem[0,0,0,0,4,4,4,4]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm10 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm6, %ymm7
-; AVX2-FAST-NEXT: vmovaps 1152(%rdi), %xmm8
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm9 = xmm8[2,3,2,3]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm6[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 768(%rdi), %xmm2
+; AVX2-FAST-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm9 = xmm2[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4],ymm10[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0,1],ymm2[2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm10, %ymm0, %ymm10
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm9 = ymm11[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm0[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm7[1,2,3],ymm2[4],ymm7[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm9 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3],ymm9[4,5,6],ymm7[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm11 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3,4],ymm10[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm12[0,1],ymm8[2,3],ymm12[4,5],ymm8[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm0, %ymm9
+; AVX2-FAST-NEXT: vmovaps 1152(%rdi), %xmm0
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm10 = xmm0[2,3,2,3]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm11 = ymm2[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm11[1,2,3],ymm10[4],ymm11[5,6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm6[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm11 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm12[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm13[1],ymm2[2,3,4],ymm13[5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1,2,3],ymm1[4],ymm9[5,6,7]
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps $34, (%rsp), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3],ymm7[4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm3[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1,2,3],ymm1[4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm9[0,1,2],ymm3[3],ymm9[4,5,6],ymm3[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm1 = mem[3,3,3,3]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm3[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm5[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1,2,3],ymm1[4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm15[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3],ymm5[4,5,6],ymm3[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm1 = mem[3,3,3,3]
-; AVX2-FAST-NEXT: vpermilps $244, (%rsp), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm14[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1,2,3],ymm1[4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm14[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm15[1],ymm3[2,3,4],ymm15[5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm8[3,3,3,3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm4[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm11[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm0[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm12[1],ymm1[2,3,4],ymm12[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm13[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1],ymm1[2,3,4],ymm6[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm9[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2,3,4],ymm8[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm11[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4],ymm15[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm4[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3,4],ymm3[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm5[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = ymm7[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm13, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm2 = [0,2,0,6,0,2,0,6]
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm3 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 272(%rdi), %xmm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm2 = [4,2,4,2]
-; AVX2-FAST-NEXT: # xmm2 = mem[0,0]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm2, %ymm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm13, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm5 = [0,2,0,6,0,2,0,6]
-; AVX2-FAST-NEXT: # ymm5 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm5, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm12[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 464(%rdi), %xmm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm2, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm8[2,3],ymm9[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm13, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm10[2,3],ymm11[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm5, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 656(%rdi), %xmm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm13, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 656(%rdi), %xmm1
+; AVX2-FAST-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm1[0,1],mem[2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm2, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm5, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm15[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm10[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 848(%rdi), %xmm1
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm2, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm12[2,3],ymm14[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm5, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 848(%rdi), %xmm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm13, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1040(%rdi), %xmm1
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm2, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm13 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm13, %ymm5, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm14 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm12 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1040(%rdi), %xmm11
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm11[2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm12, %ymm13, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm12 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 1232(%rdi), %xmm11
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1],ymm11[2,3],ymm12[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm2, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm9 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm5, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm10, %ymm2, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1232(%rdi), %xmm6
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm6[2,3],ymm9[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm13, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 1424(%rdi), %xmm8
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1],ymm8[2,3],ymm7[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm2, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm5, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm10 = ymm15[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm4 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1],ymm3[2,3],ymm10[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm5, %ymm0
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm2, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1424(%rdi), %xmm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm1[2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm13, %ymm13
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm13 = mem[0,1],ymm7[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm13, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm0 = mem[1,1,1,1,5,5,5,5]
; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm1 = [5,3,5,3]
-; AVX2-FAST-NEXT: # xmm1 = mem[0,0]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
; AVX2-FAST-NEXT: # ymm0 = mem[0,1,0,1]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, (%rsp), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm15[0,1],ymm2[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm15[0,1],ymm2[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm15[0,1],ymm2[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm5, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm13, %ymm0, %ymm13
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm12 = ymm12[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3],ymm12[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm0, %ymm9
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = ymm15[0,1,2],mem[3],ymm15[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm2[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm11[3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm12, %ymm7, %ymm11
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm11[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm10, %ymm0, %ymm10
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm6[3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm7, %ymm6
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm0, %ymm5
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm7, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm13, %ymm0, %ymm0
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, 192(%rsi)
@@ -8661,17 +8659,17 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, (%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovaps %ymm6, 224(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm9, 192(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm13, 160(%rax)
-; AVX2-FAST-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm1, 128(%rax)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm1, 96(%rax)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%rax)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm0, 224(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm5, 192(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm10, 160(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm15, 128(%rax)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm0, 96(%rax)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm0, 64(%rax)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, (%rax)
; AVX2-FAST-NEXT: addq $2504, %rsp # imm = 0x9C8
; AVX2-FAST-NEXT: vzeroupper
@@ -8679,45 +8677,47 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX2-FAST-PERLANE-LABEL: load_i32_stride6_vf64:
; AVX2-FAST-PERLANE: # %bb.0:
-; AVX2-FAST-PERLANE-NEXT: subq $2504, %rsp # imm = 0x9C8
-; AVX2-FAST-PERLANE-NEXT: vmovaps 672(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovaps 640(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT: subq $2536, %rsp # imm = 0x9E8
+; AVX2-FAST-PERLANE-NEXT: vmovaps 672(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovaps 640(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 608(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 608(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 352(%rdi), %ymm6
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %ymm7
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 256(%rdi), %ymm9
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm9, (%rsp) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm8 = <0,6,4,u>
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm8, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm8, %ymm0
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[0,1],ymm7[0,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm5 = [4,2,4,2,4,2,4,2]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm2 = [4,2,4,2,4,2,4,2]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm2, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, %ymm6
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm3[0,1],ymm2[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[0,1],ymm4[0,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 576(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm8, %ymm0
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
@@ -8728,7 +8728,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm6, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1056(%rdi), %ymm1
@@ -8753,7 +8753,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm6, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1440(%rdi), %ymm1
@@ -8778,7 +8778,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm6, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %ymm1
@@ -8786,21 +8786,21 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm8, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm15[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm8, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,2,2,2,4,6,6,6]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm6, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 480(%rdi), %ymm1
@@ -8808,21 +8808,22 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovaps 448(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 416(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 384(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm8, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm8, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,2,2,2,4,6,6,6]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 512(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 544(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm6, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, %ymm12
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 864(%rdi), %ymm1
@@ -8830,114 +8831,117 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovaps 832(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 800(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 768(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm8, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2,2,2,4,6,6,6]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 896(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 928(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm5, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 928(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm12, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm12, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm6[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1184(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1152(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1248(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1152(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm8, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1248(%rdi), %ymm9
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1216(%rdi), %ymm8
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1216(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm8[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1280(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1312(%rdi), %ymm13
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm5, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm8[0,1],ymm9[0,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1280(%rdi), %ymm9
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1312(%rdi), %ymm12
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm12[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm14, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm5 = <1,7,5,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm5, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm13 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm0 = [5,3,5,3,5,3,5,3]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm9[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm9 = [5,3,5,3,5,3,5,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm14[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm5, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm15[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm0, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm5, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm10[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm0, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm5, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm5, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm8[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm12 = ymm13[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm9, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm9, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm3[1,3,2,3,5,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm9, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm8[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm9, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %xmm4
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm4[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm5[2,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0,1],mem[2,3],ymm1[4,5],mem[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, (%rsp), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
@@ -8950,8 +8954,8 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0,1],mem[2,3],ymm1[4,5],mem[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
@@ -8966,8 +8970,8 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 960(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
@@ -8978,52 +8982,52 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2],ymm1[3],ymm5[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3],ymm4[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1344(%rdi), %xmm1
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1,2,3],ymm0[4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm15[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1,2,3],ymm0[4],ymm4[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm13[2,3],ymm5[4,5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1],ymm13[2,3],mem[4,5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,0,0,0,4,4,4,4]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm14[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm15[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm14[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3],ymm6[4,5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1,2,3],ymm0[4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1],ymm5[2,3],mem[4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1,2,3],ymm0[4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,0,0,0,4,4,4,4]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3],ymm6[4,5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 384(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -9031,99 +9035,99 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm6[1,2,3],ymm0[4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1],ymm5[2,3],mem[4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm4[0,1],mem[2,3],ymm4[4,5],mem[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm6[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3],ymm8[4,5,6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 768(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm0[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,0,0,0,4,4,4,4]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3],ymm8[4,5,6],ymm7[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1152(%rdi), %xmm12
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm7 = xmm12[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm9[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1,2,3],ymm7[4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm8[1,2,3],ymm6[4],ymm8[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm11[0,1],ymm0[2,3],ymm11[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm6[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm5[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm10 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3],ymm10[4,5,6],ymm8[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm8[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm4 = xmm4[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm7[1,2,3],ymm4[4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm3[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, (%rsp), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1152(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm8 = xmm12[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm6[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm10[1,2,3],ymm8[4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm9[0,1],ymm0[2,3],ymm9[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm10[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm4[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm11 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3],ymm8[4,5,6],ymm7[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm8[1,2,3],ymm5[4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, (%rsp), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm10 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3],ymm10[4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm8[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm3 = mem[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm5[1,2,3],ymm3[4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3],ymm7[4,5,6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm8[0,1,2],ymm5[3],ymm8[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
@@ -9138,15 +9142,14 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3],ymm5[4,5,6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm13[3,3,3,3,7,7,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
@@ -9154,302 +9157,298 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm14[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm14[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm12[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm11[1],ymm2[2,3,4],ymm11[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm9[1],ymm2[2,3,4],ymm9[5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm5[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm13[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm12[1],ymm1[2,3,4],ymm12[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1],ymm1[2,3,4],ymm6[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm7[1],ymm1[2,3,4],ymm7[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,1,0,1,4,5,4,5]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4],ymm15[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm5[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 272(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm6[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm1 = [4,2,4,2]
-; AVX2-FAST-PERLANE-NEXT: # xmm1 = mem[0,0]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm1, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm5 = [0,2,0,6,0,2,0,6]
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm5, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm4[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm14, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 464(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm8[2,3],ymm9[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm5, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 656(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vmovaps 272(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm14, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm5, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm10[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm11[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 848(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vmovaps 464(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm14, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1],ymm9[2,3],ymm10[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm12[2,3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm5, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1040(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm13 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm5, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 656(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm14, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1232(%rdi), %xmm11
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1],ymm11[2,3],ymm12[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm5, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1424(%rdi), %xmm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm5, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 848(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm14, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm10 = ymm15[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm13 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm12 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1040(%rdi), %xmm11
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1],ymm11[2,3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm14, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1],ymm3[2,3],ymm10[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm10 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1232(%rdi), %xmm6
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm6[2,3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm14, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm5, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1424(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm14, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[1,1,1,1,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm1 = [5,3,5,3]
-; AVX2-FAST-PERLANE-NEXT: # xmm1 = mem[0,0]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps (%rsp), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, (%rsp), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm0, %ymm13
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm12 = ymm12[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3],ymm12[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm13[0,1,2],ymm11[3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm9, %ymm12
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm0, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm0, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm8[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm9, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm9, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 192(%rsi)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -9531,19 +9530,19 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 224(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 192(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm13, 160(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 128(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 96(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 224(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 192(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm10, 160(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm14, 128(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 96(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 64(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, (%rax)
-; AVX2-FAST-PERLANE-NEXT: addq $2504, %rsp # imm = 0x9C8
+; AVX2-FAST-PERLANE-NEXT: addq $2536, %rsp # imm = 0x9E8
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
index 4d5c8fe891e810..a060d29200ba85 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
@@ -1779,97 +1779,101 @@ define void @load_i8_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
;
; AVX512F-LABEL: load_i8_stride5_vf32:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm0 = [65535,0,65535,65535,0,65535,0,65535,65535,0,65535,0,65535,65535,0,65535]
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm2 = [65535,65535,0,65535,0,65535,65535,0,65535,0,65535,65535,0,65535,0,65535]
; AVX512F-NEXT: vmovdqa (%rdi), %ymm3
-; AVX512F-NEXT: vmovdqa 32(%rdi), %ymm4
-; AVX512F-NEXT: vmovdqa 64(%rdi), %ymm1
-; AVX512F-NEXT: vmovdqa 96(%rdi), %ymm2
-; AVX512F-NEXT: vmovdqa %ymm0, %ymm5
-; AVX512F-NEXT: vpternlogq $202, %ymm4, %ymm3, %ymm5
-; AVX512F-NEXT: vextracti128 $1, %ymm5, %xmm6
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,zero,zero,xmm6[4,9,14],zero,zero,zero,xmm6[2,7,12,u,u,u]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[0,5,10,15],zero,zero,zero,xmm5[3,8,13],zero,zero,zero,xmm5[u,u,u]
-; AVX512F-NEXT: vpor %xmm6, %xmm5, %xmm6
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm5 = [65535,65535,0,65535,0,65535,65535,0,65535,0,65535,65535,0,65535,0,65535]
-; AVX512F-NEXT: vmovdqa %ymm5, %ymm7
-; AVX512F-NEXT: vpternlogq $202, %ymm2, %ymm1, %ymm7
-; AVX512F-NEXT: vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
-; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm7, %ymm8
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm9 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm8[1,6,11,16,21,26,31,20,25,30,19,24,29],zero,zero,zero,zero,zero,zero
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm10 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255]
-; AVX512F-NEXT: vpternlogq $248, %ymm10, %ymm6, %ymm9
+; AVX512F-NEXT: vmovdqa 32(%rdi), %ymm5
+; AVX512F-NEXT: vmovdqa 64(%rdi), %ymm0
+; AVX512F-NEXT: vmovdqa 96(%rdi), %ymm1
+; AVX512F-NEXT: vmovdqa %ymm2, %ymm4
+; AVX512F-NEXT: vpternlogq $202, %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT: vpermq {{.*#+}} ymm6 = ymm4[2,3,0,1]
+; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm6
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm8 = [128,128,128,128,128,128,128,128,128,128,128,128,128,1,6,11,16,21,26,31,20,25,30,19,24,29,128,128,128,128,128,128]
+; AVX512F-NEXT: vpshufb %ymm8, %ymm6, %ymm6
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm4 = [65535,0,65535,65535,0,65535,0,65535,65535,0,65535,0,65535,65535,0,65535]
+; AVX512F-NEXT: vmovdqa %ymm4, %ymm7
+; AVX512F-NEXT: vpternlogq $202, %ymm5, %ymm3, %ymm7
+; AVX512F-NEXT: vextracti128 $1, %ymm7, %xmm9
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm9 = zero,zero,zero,zero,xmm9[4,9,14],zero,zero,zero,xmm9[2,7,12,u,u,u]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[0,5,10,15],zero,zero,zero,xmm7[3,8,13],zero,zero,zero,xmm7[u,u,u]
+; AVX512F-NEXT: vpor %xmm7, %xmm9, %xmm9
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm11 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255]
+; AVX512F-NEXT: vpternlogq $236, %ymm11, %ymm6, %ymm9
; AVX512F-NEXT: vmovdqa 144(%rdi), %xmm7
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm6 = xmm7[u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm7[1,6,11]
+; AVX512F-NEXT: vpshufb %xmm8, %xmm7, %xmm6
; AVX512F-NEXT: vmovdqa 128(%rdi), %xmm8
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm11 = xmm8[u,u,u,u,u,u,u,u,u,u,2,7,12],zero,zero,zero
-; AVX512F-NEXT: vpor %xmm6, %xmm11, %xmm6
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm10 = xmm8[u,u,u,u,u,u,u,u,u,u,2,7,12],zero,zero,zero
+; AVX512F-NEXT: vpor %xmm6, %xmm10, %xmm6
; AVX512F-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX512F-NEXT: vpblendw {{.*#+}} ymm6 = ymm9[0,1,2,3,4],ymm6[5,6,7],ymm9[8,9,10,11,12],ymm6[13,14,15]
; AVX512F-NEXT: vpblendd {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
-; AVX512F-NEXT: vmovdqa %ymm5, %ymm9
-; AVX512F-NEXT: vpternlogq $202, %ymm4, %ymm3, %ymm9
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm11 = xmm9[1,6,11],zero,zero,zero,zero,xmm9[4,9,14],zero,zero,zero,xmm9[u,u,u]
-; AVX512F-NEXT: vextracti128 $1, %ymm9, %xmm9
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm9 = zero,zero,zero,xmm9[0,5,10,15],zero,zero,zero,xmm9[3,8,13,u,u,u]
-; AVX512F-NEXT: vpor %xmm11, %xmm9, %xmm9
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm11 = [65535,0,65535,0,65535,65535,0,65535,0,65535,65535,0,65535,0,65535,65535]
-; AVX512F-NEXT: vmovdqa %ymm11, %ymm12
-; AVX512F-NEXT: vpternlogq $202, %ymm1, %ymm2, %ymm12
-; AVX512F-NEXT: vpermq {{.*#+}} ymm13 = ymm12[2,3,0,1]
-; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm12, %ymm13
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm13[2,7,12,17,22,27,16,21,26,31,20,25,30],zero,zero,zero,zero,zero,zero
-; AVX512F-NEXT: vpternlogq $248, %ymm10, %ymm9, %ymm12
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm9 = xmm7[u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm7[2,7,12]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm13 = xmm8[u,u,u,u,u,u,u,u,u,u,3,8,13],zero,zero,zero
-; AVX512F-NEXT: vpor %xmm9, %xmm13, %xmm9
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm10 = [65535,0,65535,0,65535,65535,0,65535,0,65535,65535,0,65535,0,65535,65535]
+; AVX512F-NEXT: vmovdqa %ymm10, %ymm9
+; AVX512F-NEXT: vpternlogq $202, %ymm0, %ymm1, %ymm9
+; AVX512F-NEXT: vpermq {{.*#+}} ymm12 = ymm9[2,3,0,1]
+; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm9, %ymm12
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm9 = [128,128,128,128,128,128,128,128,128,128,128,128,128,2,7,12,17,22,27,16,21,26,31,20,25,30,128,128,128,128,128,128]
+; AVX512F-NEXT: vpshufb %ymm9, %ymm12, %ymm12
+; AVX512F-NEXT: vmovdqa %ymm2, %ymm13
+; AVX512F-NEXT: vpternlogq $202, %ymm5, %ymm3, %ymm13
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm14 = xmm13[1,6,11],zero,zero,zero,zero,xmm13[4,9,14],zero,zero,zero,xmm13[u,u,u]
+; AVX512F-NEXT: vextracti128 $1, %ymm13, %xmm13
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm13 = zero,zero,zero,xmm13[0,5,10,15],zero,zero,zero,xmm13[3,8,13,u,u,u]
+; AVX512F-NEXT: vpor %xmm14, %xmm13, %xmm13
+; AVX512F-NEXT: vpternlogq $236, %ymm11, %ymm12, %ymm13
+; AVX512F-NEXT: vpshufb %xmm9, %xmm7, %xmm9
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm12 = xmm8[u,u,u,u,u,u,u,u,u,u,3,8,13],zero,zero,zero
+; AVX512F-NEXT: vpor %xmm9, %xmm12, %xmm9
; AVX512F-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
-; AVX512F-NEXT: vpblendw {{.*#+}} ymm9 = ymm12[0,1,2,3,4],ymm9[5,6,7],ymm12[8,9,10,11,12],ymm9[13,14,15]
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm9 = ymm12[0,1,2,3],ymm9[4,5,6,7]
-; AVX512F-NEXT: vpternlogq $202, %ymm3, %ymm4, %ymm11
-; AVX512F-NEXT: vextracti128 $1, %ymm11, %xmm12
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm12 = zero,zero,zero,xmm12[1,6,11],zero,zero,zero,zero,xmm12[4,9,14,u,u,u]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[2,7,12],zero,zero,zero,xmm11[0,5,10,15],zero,zero,zero,xmm11[u,u,u]
-; AVX512F-NEXT: vpor %xmm12, %xmm11, %xmm11
-; AVX512F-NEXT: vmovdqa %ymm0, %ymm12
-; AVX512F-NEXT: vpternlogq $202, %ymm1, %ymm2, %ymm12
-; AVX512F-NEXT: vpermq {{.*#+}} ymm13 = ymm12[2,3,0,1]
-; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm12, %ymm13
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm13[3,8,13,18,23,28,17,22,27,16,21,26,31],zero,zero,zero,zero,zero,zero
-; AVX512F-NEXT: vpternlogq $248, %ymm10, %ymm11, %ymm12
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm10 = xmm7[u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm7[3,8,13]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm11 = xmm8[u,u,u,u,u,u,u,u,u,u,4,9,14],zero,zero,zero
-; AVX512F-NEXT: vpor %xmm10, %xmm11, %xmm10
-; AVX512F-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-NEXT: vpblendw {{.*#+}} ymm10 = ymm12[0,1,2,3,4],ymm10[5,6,7],ymm12[8,9,10,11,12],ymm10[13,14,15]
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
-; AVX512F-NEXT: vmovdqa %ymm0, %ymm11
-; AVX512F-NEXT: vpternlogq $202, %ymm3, %ymm4, %ymm11
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm12 = xmm11[3,8,13],zero,zero,zero,xmm11[1,6,11],zero,zero,zero,zero,xmm11[u,u,u]
-; AVX512F-NEXT: vextracti128 $1, %ymm11, %xmm11
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm11 = zero,zero,zero,xmm11[2,7,12],zero,zero,zero,xmm11[0,5,10,15,u,u,u]
-; AVX512F-NEXT: vpor %xmm12, %xmm11, %xmm11
-; AVX512F-NEXT: vmovdqa %ymm5, %ymm12
-; AVX512F-NEXT: vpternlogq $202, %ymm1, %ymm2, %ymm12
+; AVX512F-NEXT: vpblendw {{.*#+}} ymm9 = ymm13[0,1,2,3,4],ymm9[5,6,7],ymm13[8,9,10,11,12],ymm9[13,14,15]
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm9 = ymm13[0,1,2,3],ymm9[4,5,6,7]
+; AVX512F-NEXT: vmovdqa %ymm4, %ymm12
+; AVX512F-NEXT: vpternlogq $202, %ymm0, %ymm1, %ymm12
; AVX512F-NEXT: vpermq {{.*#+}} ymm13 = ymm12[2,3,0,1]
; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm12, %ymm13
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm13[4,9,14,19,24,29,18,23,28,17,22,27,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm11, %ymm12
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,u,u],zero,zero,zero,zero,xmm7[4,9,14]
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm12 = [128,128,128,128,128,128,128,128,128,128,128,128,128,3,8,13,18,23,28,17,22,27,16,21,26,31,128,128,128,128,128,128]
+; AVX512F-NEXT: vpshufb %ymm12, %ymm13, %ymm13
+; AVX512F-NEXT: vpternlogq $202, %ymm3, %ymm5, %ymm10
+; AVX512F-NEXT: vextracti128 $1, %ymm10, %xmm14
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm14 = zero,zero,zero,xmm14[1,6,11],zero,zero,zero,zero,xmm14[4,9,14,u,u,u]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[2,7,12],zero,zero,zero,xmm10[0,5,10,15],zero,zero,zero,xmm10[u,u,u]
+; AVX512F-NEXT: vpor %xmm14, %xmm10, %xmm10
+; AVX512F-NEXT: vpternlogq $236, %ymm11, %ymm13, %ymm10
+; AVX512F-NEXT: vpshufb %xmm12, %xmm7, %xmm11
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm12 = xmm8[u,u,u,u,u,u,u,u,u,u,4,9,14],zero,zero,zero
+; AVX512F-NEXT: vpor %xmm11, %xmm12, %xmm11
+; AVX512F-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
+; AVX512F-NEXT: vpblendw {{.*#+}} ymm11 = ymm10[0,1,2,3,4],ymm11[5,6,7],ymm10[8,9,10,11,12],ymm11[13,14,15]
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
+; AVX512F-NEXT: vmovdqa %ymm2, %ymm11
+; AVX512F-NEXT: vpternlogq $202, %ymm0, %ymm1, %ymm11
+; AVX512F-NEXT: vpermq {{.*#+}} ymm12 = ymm11[2,3,0,1]
+; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm11, %ymm12
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm11 = <128,128,128,128,128,128,128,128,128,128,128,128,128,4,9,14,19,24,29,18,23,28,17,22,27,u,u,u,u,u,u,u>
+; AVX512F-NEXT: vpshufb %ymm11, %ymm12, %ymm12
+; AVX512F-NEXT: vmovdqa %ymm4, %ymm13
+; AVX512F-NEXT: vpternlogq $202, %ymm3, %ymm5, %ymm13
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm14 = xmm13[3,8,13],zero,zero,zero,xmm13[1,6,11],zero,zero,zero,zero,xmm13[u,u,u]
+; AVX512F-NEXT: vextracti128 $1, %ymm13, %xmm13
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm13 = zero,zero,zero,xmm13[2,7,12],zero,zero,zero,xmm13[0,5,10,15,u,u,u]
+; AVX512F-NEXT: vpor %xmm14, %xmm13, %xmm13
+; AVX512F-NEXT: vpternlogq $236, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm12, %ymm13
+; AVX512F-NEXT: vpshufb %xmm11, %xmm7, %xmm7
; AVX512F-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[u,u,u,u,u,u,u,u,u,0,5,10,15],zero,zero,zero
; AVX512F-NEXT: vpor %xmm7, %xmm8, %xmm7
; AVX512F-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX512F-NEXT: vmovdqa {{.*#+}} ymm8 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
-; AVX512F-NEXT: vpternlogq $184, %ymm12, %ymm8, %ymm7
-; AVX512F-NEXT: vpternlogq $202, %ymm3, %ymm4, %ymm5
-; AVX512F-NEXT: vextracti128 $1, %ymm5, %xmm3
+; AVX512F-NEXT: vpternlogq $184, %ymm13, %ymm8, %ymm7
+; AVX512F-NEXT: vpternlogq $202, %ymm3, %ymm5, %ymm2
+; AVX512F-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,zero,xmm3[3,8,13],zero,zero,zero,xmm3[1,6,11,u,u,u,u]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm4 = xmm5[4,9,14],zero,zero,zero,xmm5[2,7,12],zero,zero,zero,xmm5[u,u,u,u]
-; AVX512F-NEXT: vpor %xmm3, %xmm4, %xmm3
-; AVX512F-NEXT: vpternlogq $202, %ymm2, %ymm1, %ymm0
-; AVX512F-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm1
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm0 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,0,5,10,15,20,25,30,19,24,29,18,23,28,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[4,9,14],zero,zero,zero,xmm2[2,7,12],zero,zero,zero,xmm2[u,u,u,u]
+; AVX512F-NEXT: vpor %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT: vpternlogq $202, %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT: vpermq {{.*#+}} ymm0 = ymm4[2,3,0,1]
+; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm0
+; AVX512F-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,0,5,10,15,20,25,30,19,24,29,18,23,28,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX512F-NEXT: vmovdqa 128(%rdi), %ymm1
; AVX512F-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,1,6,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,21,26,31,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm2 = [0,5,0,5,0,5,0,5]
@@ -3821,202 +3825,206 @@ define void @load_i8_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
;
; AVX512F-LABEL: load_i8_stride5_vf64:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm0 = [65535,0,65535,65535,0,65535,0,65535,65535,0,65535,0,65535,65535,0,65535]
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm5 = [65535,65535,0,65535,0,65535,65535,0,65535,0,65535,65535,0,65535,0,65535]
; AVX512F-NEXT: vmovdqa64 (%rdi), %ymm23
; AVX512F-NEXT: vmovdqa64 32(%rdi), %ymm24
; AVX512F-NEXT: vmovdqa64 64(%rdi), %ymm21
; AVX512F-NEXT: vmovdqa64 96(%rdi), %ymm22
-; AVX512F-NEXT: vmovdqa %ymm0, %ymm5
-; AVX512F-NEXT: vpternlogq $202, %ymm24, %ymm23, %ymm5
-; AVX512F-NEXT: vextracti128 $1, %ymm5, %xmm6
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,zero,zero,xmm6[4,9,14],zero,zero,zero,xmm6[2,7,12,u,u,u]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[0,5,10,15],zero,zero,zero,xmm5[3,8,13],zero,zero,zero,xmm5[u,u,u]
-; AVX512F-NEXT: vpor %xmm6, %xmm5, %xmm6
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm5 = [65535,65535,0,65535,0,65535,65535,0,65535,0,65535,65535,0,65535,0,65535]
-; AVX512F-NEXT: vmovdqa %ymm5, %ymm7
-; AVX512F-NEXT: vpternlogq $202, %ymm22, %ymm21, %ymm7
-; AVX512F-NEXT: vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
-; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm7, %ymm8
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm9 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm8[1,6,11,16,21,26,31,20,25,30,19,24,29],zero,zero,zero,zero,zero,zero
-; AVX512F-NEXT: vmovdqa64 {{.*#+}} ymm18 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255]
-; AVX512F-NEXT: vpternlogq $248, %ymm18, %ymm6, %ymm9
-; AVX512F-NEXT: vmovdqa 192(%rdi), %ymm6
+; AVX512F-NEXT: vmovdqa %ymm5, %ymm4
+; AVX512F-NEXT: vpternlogq $202, %ymm22, %ymm21, %ymm4
+; AVX512F-NEXT: vpermq {{.*#+}} ymm6 = ymm4[2,3,0,1]
+; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm6
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm10 = [128,128,128,128,128,128,128,128,128,128,128,128,128,1,6,11,16,21,26,31,20,25,30,19,24,29,128,128,128,128,128,128]
+; AVX512F-NEXT: vpshufb %ymm10, %ymm6, %ymm6
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm4 = [65535,0,65535,65535,0,65535,0,65535,65535,0,65535,0,65535,65535,0,65535]
+; AVX512F-NEXT: vmovdqa %ymm4, %ymm7
+; AVX512F-NEXT: vpternlogq $202, %ymm24, %ymm23, %ymm7
+; AVX512F-NEXT: vextracti128 $1, %ymm7, %xmm8
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm8 = zero,zero,zero,zero,xmm8[4,9,14],zero,zero,zero,xmm8[2,7,12,u,u,u]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[0,5,10,15],zero,zero,zero,xmm7[3,8,13],zero,zero,zero,xmm7[u,u,u]
+; AVX512F-NEXT: vpor %xmm7, %xmm8, %xmm12
+; AVX512F-NEXT: vmovdqa64 {{.*#+}} ymm19 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255]
+; AVX512F-NEXT: vpternlogq $236, %ymm19, %ymm6, %ymm12
+; AVX512F-NEXT: vmovdqa64 192(%rdi), %ymm25
; AVX512F-NEXT: vmovdqa 224(%rdi), %ymm7
-; AVX512F-NEXT: vmovdqa %ymm0, %ymm10
-; AVX512F-NEXT: vpternlogq $202, %ymm6, %ymm7, %ymm10
+; AVX512F-NEXT: vmovdqa %ymm4, %ymm9
+; AVX512F-NEXT: vpternlogq $202, %ymm25, %ymm7, %ymm9
; AVX512F-NEXT: vmovdqa 208(%rdi), %xmm8
-; AVX512F-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm10
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,zero,zero,zero,zero,zero,ymm10[3,8,13,2,7,12,1,6,11,16,21,26,31,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vmovdqa 176(%rdi), %xmm10
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm13 = xmm10[u,u,u,u,u,u,u,u,4,9,14,u,u,u,u,u]
+; AVX512F-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm9
+; AVX512F-NEXT: vpshufb {{.*#+}} ymm13 = zero,zero,zero,zero,zero,zero,zero,ymm9[3,8,13,2,7,12,1,6,11,16,21,26,31,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vmovdqa 176(%rdi), %xmm9
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm14 = xmm9[u,u,u,u,u,u,u,u,4,9,14,u,u,u,u,u]
; AVX512F-NEXT: vmovdqa 160(%rdi), %xmm11
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm14 = xmm11[u,u,u,u,u,u,u,u,0,5,10,15,u,u,u,u]
-; AVX512F-NEXT: vpunpckhdq {{.*#+}} xmm1 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm15 = xmm11[u,u,u,u,u,u,u,u,0,5,10,15,u,u,u,u]
+; AVX512F-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm15[2],xmm14[2],xmm15[3],xmm14[3]
; AVX512F-NEXT: vmovdqa64 {{.*#+}} ymm16 = [0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512F-NEXT: vpternlogq $186, %ymm12, %ymm16, %ymm1
+; AVX512F-NEXT: vpternlogq $186, %ymm13, %ymm16, %ymm0
; AVX512F-NEXT: vmovdqa 144(%rdi), %xmm13
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm12 = xmm13[u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm13[1,6,11]
+; AVX512F-NEXT: vpshufb %xmm10, %xmm13, %xmm10
; AVX512F-NEXT: vmovdqa 128(%rdi), %xmm14
; AVX512F-NEXT: vpshufb {{.*#+}} xmm15 = xmm14[u,u,u,u,u,u,u,u,u,u,2,7,12],zero,zero,zero
-; AVX512F-NEXT: vpor %xmm12, %xmm15, %xmm12
-; AVX512F-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
-; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm12, %zmm17
+; AVX512F-NEXT: vpor %xmm10, %xmm15, %xmm10
+; AVX512F-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
+; AVX512F-NEXT: vinserti64x4 $1, %ymm0, %zmm10, %zmm10
; AVX512F-NEXT: vmovdqa64 {{.*#+}} zmm20 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
-; AVX512F-NEXT: vpternlogq $184, %zmm9, %zmm20, %zmm17
+; AVX512F-NEXT: vpternlogq $184, %zmm12, %zmm20, %zmm10
; AVX512F-NEXT: vmovdqa 256(%rdi), %ymm15
; AVX512F-NEXT: vmovdqa 288(%rdi), %ymm12
-; AVX512F-NEXT: vmovdqa %ymm5, %ymm9
-; AVX512F-NEXT: vpternlogq $202, %ymm15, %ymm12, %ymm9
-; AVX512F-NEXT: vextracti128 $1, %ymm9, %xmm2
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u],zero,zero,zero,xmm2[3,8,13],zero,zero,zero,xmm2[1,6,11]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm9 = xmm9[u,u,u,u,4,9,14],zero,zero,zero,xmm9[2,7,12],zero,zero,zero
-; AVX512F-NEXT: vpor %xmm2, %xmm9, %xmm2
+; AVX512F-NEXT: vmovdqa %ymm5, %ymm2
+; AVX512F-NEXT: vpternlogq $202, %ymm15, %ymm12, %ymm2
+; AVX512F-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u],zero,zero,zero,xmm3[3,8,13],zero,zero,zero,xmm3[1,6,11]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,4,9,14],zero,zero,zero,xmm2[2,7,12],zero,zero,zero
+; AVX512F-NEXT: vpor %xmm3, %xmm2, %xmm2
; AVX512F-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm17, %zmm19
-; AVX512F-NEXT: vmovdqa %ymm0, %ymm1
-; AVX512F-NEXT: vpternlogq $202, %ymm12, %ymm15, %ymm1
-; AVX512F-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX512F-NEXT: vinserti64x4 $1, %ymm0, %zmm10, %zmm18
+; AVX512F-NEXT: vmovdqa %ymm4, %ymm0
+; AVX512F-NEXT: vpternlogq $202, %ymm12, %ymm15, %ymm0
+; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u],zero,zero,zero,zero,xmm2[4,9,14],zero,zero,zero,xmm2[2,7,12]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,0,5,10,15],zero,zero,zero,xmm1[3,8,13],zero,zero,zero
-; AVX512F-NEXT: vpor %xmm2, %xmm1, %xmm1
-; AVX512F-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,0,5,10,15],zero,zero,zero,xmm0[3,8,13],zero,zero,zero
+; AVX512F-NEXT: vpor %xmm2, %xmm0, %xmm0
+; AVX512F-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512F-NEXT: vmovdqa %ymm5, %ymm2
-; AVX512F-NEXT: vpternlogq $202, %ymm6, %ymm7, %ymm2
+; AVX512F-NEXT: vpternlogq $202, %ymm25, %ymm7, %ymm2
; AVX512F-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm2
; AVX512F-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,ymm2[4,9,14,3,8,13,2,7,12,17,22,27,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm9 = xmm11[1,6,11],zero,zero,zero,zero,xmm11[u,u,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,zero,xmm10[0,5,10,15,u,u,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpor %xmm3, %xmm9, %xmm3
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = xmm11[1,6,11],zero,zero,zero,zero,xmm11[u,u,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm10 = zero,zero,zero,xmm9[0,5,10,15,u,u,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpor %xmm3, %xmm10, %xmm3
; AVX512F-NEXT: vpternlogq $186, %ymm2, %ymm16, %ymm3
; AVX512F-NEXT: vmovdqa64 {{.*#+}} ymm16 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX512F-NEXT: vpternlogq $226, %ymm1, %ymm16, %ymm3
+; AVX512F-NEXT: vpternlogq $226, %ymm0, %ymm16, %ymm3
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm10 = [65535,0,65535,0,65535,65535,0,65535,0,65535,65535,0,65535,0,65535,65535]
+; AVX512F-NEXT: vmovdqa %ymm10, %ymm0
+; AVX512F-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm0
+; AVX512F-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm0 = [128,128,128,128,128,128,128,128,128,128,128,128,128,2,7,12,17,22,27,16,21,26,31,20,25,30,128,128,128,128,128,128]
+; AVX512F-NEXT: vpshufb %ymm0, %ymm2, %ymm2
; AVX512F-NEXT: vmovdqa %ymm5, %ymm1
; AVX512F-NEXT: vpternlogq $202, %ymm24, %ymm23, %ymm1
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = xmm1[1,6,11],zero,zero,zero,zero,xmm1[4,9,14],zero,zero,zero,xmm1[u,u,u]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm6 = xmm1[1,6,11],zero,zero,zero,zero,xmm1[4,9,14],zero,zero,zero,xmm1[u,u,u]
; AVX512F-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,zero,xmm1[0,5,10,15],zero,zero,zero,xmm1[3,8,13,u,u,u]
-; AVX512F-NEXT: vpor %xmm2, %xmm1, %xmm1
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm9 = [65535,0,65535,0,65535,65535,0,65535,0,65535,65535,0,65535,0,65535,65535]
-; AVX512F-NEXT: vmovdqa %ymm9, %ymm2
-; AVX512F-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm2
-; AVX512F-NEXT: vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
-; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm4
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm4[2,7,12,17,22,27,16,21,26,31,20,25,30],zero,zero,zero,zero,zero,zero
-; AVX512F-NEXT: vpternlogq $248, %ymm18, %ymm1, %ymm2
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm13[u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm13[2,7,12]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm4 = xmm14[u,u,u,u,u,u,u,u,u,u,3,8,13],zero,zero,zero
-; AVX512F-NEXT: vpor %xmm1, %xmm4, %xmm1
-; AVX512F-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-NEXT: vpternlogq $184, %zmm2, %zmm20, %zmm1
-; AVX512F-NEXT: vinserti64x4 $1, %ymm3, %zmm1, %zmm17
+; AVX512F-NEXT: vpor %xmm6, %xmm1, %xmm1
+; AVX512F-NEXT: vpternlogq $236, %ymm19, %ymm2, %ymm1
+; AVX512F-NEXT: vpshufb %xmm0, %xmm13, %xmm0
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = xmm14[u,u,u,u,u,u,u,u,u,u,3,8,13],zero,zero,zero
+; AVX512F-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX512F-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-NEXT: vpternlogq $184, %zmm1, %zmm20, %zmm0
+; AVX512F-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm17
+; AVX512F-NEXT: vmovdqa %ymm5, %ymm0
+; AVX512F-NEXT: vpternlogq $202, %ymm12, %ymm15, %ymm0
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm0[u,u,u,1,6,11],zero,zero,zero,zero,xmm0[4,9,14],zero,zero,zero
+; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u],zero,zero,zero,xmm0[0,5,10,15],zero,zero,zero,xmm0[3,8,13]
+; AVX512F-NEXT: vpor %xmm1, %xmm0, %xmm0
+; AVX512F-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-NEXT: vmovdqa %ymm4, %ymm1
+; AVX512F-NEXT: vpternlogq $202, %ymm7, %ymm25, %ymm1
+; AVX512F-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm1
+; AVX512F-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,0,5,10,15,4,9,14,3,8,13,18,23,28,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm9[1,6,11,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = xmm11[2,7,12],zero,zero,zero,xmm11[u,u,u,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpor %xmm2, %xmm3, %xmm2
+; AVX512F-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm1[3,4,5,6,7]
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT: vpternlogq $226, %ymm0, %ymm16, %ymm1
+; AVX512F-NEXT: vmovdqa %ymm4, %ymm0
+; AVX512F-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm0
+; AVX512F-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm0 = [128,128,128,128,128,128,128,128,128,128,128,128,128,3,8,13,18,23,28,17,22,27,16,21,26,31,128,128,128,128,128,128]
+; AVX512F-NEXT: vpshufb %ymm0, %ymm2, %ymm2
+; AVX512F-NEXT: vmovdqa %ymm10, %ymm3
+; AVX512F-NEXT: vpternlogq $202, %ymm23, %ymm24, %ymm3
+; AVX512F-NEXT: vextracti128 $1, %ymm3, %xmm6
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,zero,xmm6[1,6,11],zero,zero,zero,zero,xmm6[4,9,14,u,u,u]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[2,7,12],zero,zero,zero,xmm3[0,5,10,15],zero,zero,zero,xmm3[u,u,u]
+; AVX512F-NEXT: vpor %xmm6, %xmm3, %xmm3
+; AVX512F-NEXT: vpternlogq $236, %ymm19, %ymm2, %ymm3
+; AVX512F-NEXT: vpshufb %xmm0, %xmm13, %xmm0
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = xmm14[u,u,u,u,u,u,u,u,u,u,4,9,14],zero,zero,zero
+; AVX512F-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX512F-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-NEXT: vpternlogq $184, %zmm3, %zmm20, %zmm0
+; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm19
+; AVX512F-NEXT: vmovdqa %ymm10, %ymm0
+; AVX512F-NEXT: vpternlogq $202, %ymm15, %ymm12, %ymm0
+; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u],zero,zero,zero,xmm1[1,6,11],zero,zero,zero,zero,xmm1[4,9,14]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,2,7,12],zero,zero,zero,xmm0[0,5,10,15],zero,zero,zero
+; AVX512F-NEXT: vpor %xmm1, %xmm0, %xmm0
+; AVX512F-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512F-NEXT: vmovdqa %ymm5, %ymm1
-; AVX512F-NEXT: vpternlogq $202, %ymm12, %ymm15, %ymm1
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = xmm1[u,u,u,1,6,11],zero,zero,zero,zero,xmm1[4,9,14],zero,zero,zero
-; AVX512F-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u],zero,zero,zero,xmm1[0,5,10,15],zero,zero,zero,xmm1[3,8,13]
-; AVX512F-NEXT: vpor %xmm2, %xmm1, %xmm1
-; AVX512F-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-NEXT: vmovdqa %ymm0, %ymm2
-; AVX512F-NEXT: vpternlogq $202, %ymm7, %ymm6, %ymm2
-; AVX512F-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm2
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,0,5,10,15,4,9,14,3,8,13,18,23,28,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,zero,xmm10[1,6,11,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm4 = xmm11[2,7,12],zero,zero,zero,xmm11[u,u,u,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpor %xmm3, %xmm4, %xmm3
-; AVX512F-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm2[3,4,5,6,7]
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-NEXT: vpternlogq $226, %ymm1, %ymm16, %ymm2
-; AVX512F-NEXT: vmovdqa %ymm9, %ymm1
-; AVX512F-NEXT: vpternlogq $202, %ymm23, %ymm24, %ymm1
-; AVX512F-NEXT: vextracti128 $1, %ymm1, %xmm3
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,zero,xmm3[1,6,11],zero,zero,zero,zero,xmm3[4,9,14,u,u,u]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[2,7,12],zero,zero,zero,xmm1[0,5,10,15],zero,zero,zero,xmm1[u,u,u]
-; AVX512F-NEXT: vpor %xmm3, %xmm1, %xmm1
-; AVX512F-NEXT: vmovdqa %ymm0, %ymm3
-; AVX512F-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm3
-; AVX512F-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
-; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm4
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm4[3,8,13,18,23,28,17,22,27,16,21,26,31],zero,zero,zero,zero,zero,zero
-; AVX512F-NEXT: vpternlogq $248, %ymm18, %ymm1, %ymm3
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm13[u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm13[3,8,13]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm4 = xmm14[u,u,u,u,u,u,u,u,u,u,4,9,14],zero,zero,zero
-; AVX512F-NEXT: vpor %xmm1, %xmm4, %xmm1
-; AVX512F-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-NEXT: vpternlogq $184, %zmm3, %zmm20, %zmm1
-; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm18
-; AVX512F-NEXT: vmovdqa %ymm9, %ymm1
-; AVX512F-NEXT: vpternlogq $202, %ymm15, %ymm12, %ymm1
-; AVX512F-NEXT: vextracti128 $1, %ymm1, %xmm2
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u],zero,zero,zero,xmm2[1,6,11],zero,zero,zero,zero,xmm2[4,9,14]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,2,7,12],zero,zero,zero,xmm1[0,5,10,15],zero,zero,zero
-; AVX512F-NEXT: vpor %xmm2, %xmm1, %xmm1
-; AVX512F-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-NEXT: vmovdqa %ymm5, %ymm2
-; AVX512F-NEXT: vpternlogq $202, %ymm7, %ymm6, %ymm2
-; AVX512F-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm2
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,1,6,11,0,5,10,15,4,9,14,19,24,29,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,zero,xmm10[2,7,12,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm4 = xmm11[3,8,13],zero,zero,zero,xmm11[u,u,u,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpor %xmm3, %xmm4, %xmm3
-; AVX512F-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm2[3,4,5,6,7]
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-NEXT: vpternlogq $226, %ymm1, %ymm16, %ymm2
-; AVX512F-NEXT: vmovdqa %ymm0, %ymm1
-; AVX512F-NEXT: vpternlogq $202, %ymm23, %ymm24, %ymm1
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = xmm1[3,8,13],zero,zero,zero,xmm1[1,6,11],zero,zero,zero,zero,xmm1[u,u,u]
-; AVX512F-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,zero,xmm1[2,7,12],zero,zero,zero,xmm1[0,5,10,15,u,u,u]
-; AVX512F-NEXT: vpor %xmm3, %xmm1, %xmm1
-; AVX512F-NEXT: vmovdqa %ymm5, %ymm3
-; AVX512F-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm3
-; AVX512F-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
-; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm4
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm4[4,9,14,19,24,29,18,23,28,17,22,27,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm3
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm13[u,u,u,u,u,u,u,u,u],zero,zero,zero,zero,xmm13[4,9,14]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm4 = xmm14[u,u,u,u,u,u,u,u,u,0,5,10,15],zero,zero,zero
-; AVX512F-NEXT: vpor %xmm1, %xmm4, %xmm1
-; AVX512F-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-NEXT: vmovdqa64 {{.*#+}} zmm4 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512F-NEXT: vpternlogq $226, %zmm3, %zmm4, %zmm1
-; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512F-NEXT: vpternlogq $226, %ymm15, %ymm0, %ymm12
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = xmm12[u,u,u,3,8,13],zero,zero,zero,xmm12[1,6,11],zero,zero,zero,zero
+; AVX512F-NEXT: vpternlogq $202, %ymm7, %ymm25, %ymm1
+; AVX512F-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm1
+; AVX512F-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,1,6,11,0,5,10,15,4,9,14,19,24,29,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm9[2,7,12,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = xmm11[3,8,13],zero,zero,zero,xmm11[u,u,u,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpor %xmm2, %xmm3, %xmm2
+; AVX512F-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm1[3,4,5,6,7]
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-NEXT: vpternlogq $226, %ymm0, %ymm16, %ymm1
+; AVX512F-NEXT: vmovdqa %ymm5, %ymm0
+; AVX512F-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm0
+; AVX512F-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm0 = <128,128,128,128,128,128,128,128,128,128,128,128,128,4,9,14,19,24,29,18,23,28,17,22,27,u,u,u,u,u,u,u>
+; AVX512F-NEXT: vpshufb %ymm0, %ymm2, %ymm2
+; AVX512F-NEXT: vmovdqa %ymm4, %ymm3
+; AVX512F-NEXT: vpternlogq $202, %ymm23, %ymm24, %ymm3
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm6 = xmm3[3,8,13],zero,zero,zero,xmm3[1,6,11],zero,zero,zero,zero,xmm3[u,u,u]
+; AVX512F-NEXT: vextracti128 $1, %ymm3, %xmm3
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,zero,xmm3[2,7,12],zero,zero,zero,xmm3[0,5,10,15,u,u,u]
+; AVX512F-NEXT: vpor %xmm6, %xmm3, %xmm3
+; AVX512F-NEXT: vpternlogq $236, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm3
+; AVX512F-NEXT: vpshufb %xmm0, %xmm13, %xmm0
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = xmm14[u,u,u,u,u,u,u,u,u,0,5,10,15],zero,zero,zero
+; AVX512F-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX512F-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-NEXT: vpternlogq $226, %zmm3, %zmm2, %zmm0
+; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT: vpternlogq $226, %ymm15, %ymm4, %ymm12
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm12[u,u,u,3,8,13],zero,zero,zero,xmm12[1,6,11],zero,zero,zero,zero
; AVX512F-NEXT: vextracti128 $1, %ymm12, %xmm3
; AVX512F-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u],zero,zero,zero,xmm3[2,7,12],zero,zero,zero,xmm3[0,5,10,15]
-; AVX512F-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX512F-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512F-NEXT: vpternlogq $202, %ymm6, %ymm7, %ymm9
-; AVX512F-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm9
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm3 = ymm9[u,u,u,u,u,u,2,7,12,1,6,11,0,5,10,15,20,25,30,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,zero,xmm10[3,8,13,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpor %xmm1, %xmm3, %xmm1
+; AVX512F-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512F-NEXT: vpternlogq $202, %ymm25, %ymm7, %ymm10
+; AVX512F-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm10
+; AVX512F-NEXT: vpshufb {{.*#+}} ymm3 = ymm10[u,u,u,u,u,u,2,7,12,1,6,11,0,5,10,15,20,25,30,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,zero,xmm9[3,8,13,u,u,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vpshufb {{.*#+}} xmm7 = xmm11[4,9,14],zero,zero,zero,xmm11[u,u,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vpor %xmm6, %xmm7, %xmm6
; AVX512F-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2],xmm3[3,4,5,6,7]
; AVX512F-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX512F-NEXT: vpternlogq $226, %ymm2, %ymm16, %ymm3
+; AVX512F-NEXT: vpternlogq $226, %ymm1, %ymm16, %ymm3
; AVX512F-NEXT: vpternlogq $202, %ymm23, %ymm24, %ymm5
-; AVX512F-NEXT: vextracti128 $1, %ymm5, %xmm2
-; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm2[3,8,13],zero,zero,zero,xmm2[1,6,11,u,u,u,u]
+; AVX512F-NEXT: vextracti128 $1, %ymm5, %xmm1
+; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,zero,xmm1[3,8,13],zero,zero,zero,xmm1[1,6,11,u,u,u,u]
; AVX512F-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[4,9,14],zero,zero,zero,xmm5[2,7,12],zero,zero,zero,xmm5[u,u,u,u]
-; AVX512F-NEXT: vpor %xmm2, %xmm5, %xmm2
-; AVX512F-NEXT: vpternlogq $202, %ymm22, %ymm21, %ymm0
-; AVX512F-NEXT: vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
-; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm5
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm0 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,0,5,10,15,20,25,30,19,24,29,18,23,28,u,u,u,u,u,u,u]
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
-; AVX512F-NEXT: vmovdqa 128(%rdi), %ymm2
-; AVX512F-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,1,6,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,21,26,31,u,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpor %xmm1, %xmm5, %xmm1
+; AVX512F-NEXT: vpternlogq $202, %ymm22, %ymm21, %ymm4
+; AVX512F-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
+; AVX512F-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm5
+; AVX512F-NEXT: vpshufb {{.*#+}} ymm4 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,0,5,10,15,20,25,30,19,24,29,18,23,28,u,u,u,u,u,u,u]
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5,6,7]
+; AVX512F-NEXT: vmovdqa 128(%rdi), %ymm4
+; AVX512F-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[u,1,6,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,21,26,31,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm5 = [0,5,0,5,0,5,0,5]
-; AVX512F-NEXT: vpermd %ymm2, %ymm5, %ymm2
-; AVX512F-NEXT: vpternlogq $226, %zmm0, %zmm4, %zmm2
-; AVX512F-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm0
-; AVX512F-NEXT: vmovdqa64 %zmm19, (%rsi)
+; AVX512F-NEXT: vpermd %ymm4, %ymm5, %ymm4
+; AVX512F-NEXT: vpternlogq $226, %zmm1, %zmm2, %zmm4
+; AVX512F-NEXT: vinserti64x4 $1, %ymm3, %zmm4, %zmm1
+; AVX512F-NEXT: vmovdqa64 %zmm18, (%rsi)
; AVX512F-NEXT: vmovdqa64 %zmm17, (%rdx)
-; AVX512F-NEXT: vmovdqa64 %zmm18, (%rcx)
-; AVX512F-NEXT: vmovdqa64 %zmm1, (%r8)
-; AVX512F-NEXT: vmovdqa64 %zmm0, (%r9)
+; AVX512F-NEXT: vmovdqa64 %zmm19, (%rcx)
+; AVX512F-NEXT: vmovdqa64 %zmm0, (%r8)
+; AVX512F-NEXT: vmovdqa64 %zmm1, (%r9)
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
index be34bbd2540222..992b190ac17cc8 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
@@ -1700,248 +1700,250 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
;
; AVX1-ONLY-LABEL: load_i8_stride6_vf32:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $168, %rsp
-; AVX1-ONLY-NEXT: vmovdqa 128(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm1[u,u,4,10,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa 144(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm2[u,u,u,u,0,6,12,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm5
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm2[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm2[4,10]
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm6
-; AVX1-ONLY-NEXT: vmovdqa 160(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm3[u,u,u,u,u,u,u,u,u,u,u,2,8,14],zero,zero
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm7
+; AVX1-ONLY-NEXT: subq $120, %rsp
+; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vmovdqa 32(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovdqa 48(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm8[2,8,14,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm6[u,u,u,u,u,0,6,12,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = zero,zero,zero,xmm7[4,10,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm9[2,8,14],zero,zero,xmm9[u,u,u,u,u,u,u,u,u,u,u]
; AVX1-ONLY-NEXT: vpor %xmm0, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,u,u,255,255,255,255,255,0,0,0,0,0>
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = <0,0,0,0,0,255,255,255,255,255,255,u,u,u,u,u>
; AVX1-ONLY-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm4[u,u,5,11,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm5[u,u,u,u,1,7,13,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm6[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm6[5,11]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,u,u,u,u,3,9,15],zero,zero
+; AVX1-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm8[3,9,15,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm6[u,u,u,u,u,1,7,13,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm7[5,11,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm9[3,9,15],zero,zero,xmm9[u,u,u,u,u,u,u,u,u,u,u]
; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
; AVX1-ONLY-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 32(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vmovdqa 48(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm0 = [2,8,14,0,2,8,14,0,2,8,14,0,2,8,14,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm15, %xmm2
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm1 = [0,0,6,12,0,0,6,12,0,0,6,12,0,0,6,12]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm14, %xmm3
-; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm9
-; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm8[4,10,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm10 = xmm9[2,8,14],zero,zero,xmm9[u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm10, %xmm10
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm2 = <0,0,0,0,0,255,255,255,255,255,255,u,u,u,u,u>
-; AVX1-ONLY-NEXT: vpblendvb %xmm2, %xmm3, %xmm10, %xmm3
-; AVX1-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [3,9,15,0,3,9,15,0,3,9,15,0,3,9,15,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm15, %xmm10
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm11 = [0,1,7,13,0,1,7,13,0,1,7,13,0,1,7,13]
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm14, %xmm12
-; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm10 = xmm12[0],xmm10[0]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = zero,zero,zero,xmm8[5,11,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm9[3,9,15],zero,zero,xmm9[u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm12, %xmm13, %xmm12
-; AVX1-ONLY-NEXT: vpblendvb %xmm2, %xmm10, %xmm12, %xmm2
-; AVX1-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm5, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm4, %xmm1
-; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm7[u,u,u,u,u,u,u,u,u,u,u,4,10],zero,zero,zero
-; AVX1-ONLY-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm6[u,u,u,u,u,u,u,u,u,u,u],zero,zero,xmm6[0,6,12]
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,u,u,u,255,255,255,255,255,255,0,0,0,0,0>
-; AVX1-ONLY-NEXT: vpblendvb %xmm2, %xmm0, %xmm1, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm5, %xmm0
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm4, %xmm1
-; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm7[u,u,u,u,u,u,u,u,u,u,u,5,11],zero,zero,zero
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm6[u,u,u,u,u,u,u,u,u,u,u],zero,zero,xmm6[1,7,13]
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm3, %xmm1
-; AVX1-ONLY-NEXT: vpblendvb %xmm2, %xmm0, %xmm1, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm9[4,10],zero,zero,zero,xmm9[u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,xmm8[0,6,12,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,xmm7[0,6,12,u,u,u,u,u,u,u,u,u,u,u]
; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm15[u,u,u,u,u,u,u,u,4,10,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm14[u,u,u,u,u,u,u,u,u,u,u,u,u,2,8,14]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm8[u,u,u,u,u,u,u,u,4,10,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm6[u,u,u,u,u,u,u,u,u,u,u,u,u,2,8,14]
; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = <255,255,255,255,255,0,0,0,0,0,u,u,u,u,u,u>
; AVX1-ONLY-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm1
; AVX1-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm9[5,11],zero,zero,zero,xmm9[u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,xmm8[1,7,13,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,xmm7[1,7,13,u,u,u,u,u,u,u,u,u,u,u]
; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm15[u,u,u,u,u,u,u,u,5,11,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm14[u,u,u,u,u,u,u,u,u,u,u,u,u,3,9,15]
+; AVX1-ONLY-NEXT: vmovdqa %xmm8, %xmm5
+; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm8[u,u,u,u,u,u,u,u,5,11,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm6[u,u,u,u,u,u,u,u,u,u,u,u,u,3,9,15]
; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm0
; AVX1-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm10 = [128,128,128,2,8,14,0,0,128,128,128,2,8,14,0,0]
-; AVX1-ONLY-NEXT: # xmm10 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa 112(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm5, %xmm1
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = [0,6,12,128,128,128,0,0,0,6,12,128,128,128,0,0]
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa 80(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm2[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm2[4,10]
-; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm4[u,u,u,u,u,u,u,u,u,u,u,2,8,14],zero,zero
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm13, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm8 = [128,128,128,2,8,14,0,0,128,128,128,2,8,14,0,0]
+; AVX1-ONLY-NEXT: # xmm8 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa 112(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm0, %xmm1
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [0,6,12,128,128,128,0,0,0,6,12,128,128,128,0,0]
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm13, %xmm3
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm3, %xmm1
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [0,0,0,128,128,128,4,10,0,0,0,128,128,128,4,10]
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa 80(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm12, %xmm4
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [0,0,0,2,8,14,128,128,0,0,0,2,8,14,128,128]
+; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm14, %xmm10
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm10, %xmm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm4, %ymm1
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm15 = [0,0,4,10,0,0,4,10,0,0,4,10,0,0,4,10]
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm6, %xmm4
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm0 = [0,6,12,0,0,6,12,0,0,6,12,0,0,6,12,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm5, %xmm10
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm10[0],xmm4[0],xmm10[1],xmm4[1]
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm7, %xmm8
+; AVX1-ONLY-NEXT: vmovdqa %xmm7, %xmm10
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm9, %xmm2
+; AVX1-ONLY-NEXT: vpor %xmm2, %xmm8, %xmm2
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm4[3,4,5],xmm2[6,7]
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255]
+; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm4, %ymm1
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm2, %ymm8
+; AVX1-ONLY-NEXT: vmovdqa 128(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm6, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa 144(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm5, %xmm0
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm15 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm4, %xmm11
+; AVX1-ONLY-NEXT: vmovdqa 160(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm2, %xmm3
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255]
+; AVX1-ONLY-NEXT: vpblendvb %xmm3, %xmm15, %xmm11, %xmm15
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm11 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandps %ymm11, %ymm8, %ymm8
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX1-ONLY-NEXT: vandnps %ymm15, %ymm11, %ymm15
+; AVX1-ONLY-NEXT: vorps %ymm15, %ymm8, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = zero,zero,zero,xmm9[3,9,15,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm15 = xmm13[1,7,13],zero,zero,zero,xmm13[u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm15, %xmm1
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm12[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm12[5,11]
+; AVX1-ONLY-NEXT: vmovdqa %xmm14, %xmm7
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm15 = xmm14[u,u,u,u,u,u,u,u,u,u,u,3,9,15],zero,zero
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm15, %xmm0
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm14[u,u,4,10,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm15[u,u,u,u,0,6,12,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm8, %xmm10
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm9, %xmm6
-; AVX1-ONLY-NEXT: vpor %xmm6, %xmm10, %xmm6
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm6[0,1,2],xmm1[3,4,5],xmm6[6,7]
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm6 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255]
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm6, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm6, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm10, %ymm1, %ymm10
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm8 = [0,0,5,11,0,0,5,11,0,0,5,11,0,0,5,11]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm1, %xmm15
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm1 = [1,7,13,0,1,7,13,0,1,7,13,0,1,7,13,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm14, %xmm14
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm15 = zero,zero,zero,xmm10[3,9,15,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[1,7,13],zero,zero,zero,xmm10[u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm15, %xmm10, %xmm10
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2],xmm14[3,4,5],xmm10[6,7]
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm14 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255]
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm14, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm14, %ymm10, %ymm10
; AVX1-ONLY-NEXT: vorps %ymm0, %ymm10, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [128,128,128,3,9,15,0,0,128,128,128,3,9,15,0,0]
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm5, %xmm13
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [1,7,13,128,128,128,0,0,1,7,13,128,128,128,0,0]
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm3, %xmm12
-; AVX1-ONLY-NEXT: vpor %xmm13, %xmm12, %xmm12
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm2[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm2[5,11]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm4[u,u,u,u,u,u,u,u,u,u,u,3,9,15],zero,zero
-; AVX1-ONLY-NEXT: vpor %xmm13, %xmm11, %xmm11
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm11, %ymm11
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm14[u,u,5,11,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm15[u,u,u,u,1,7,13,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm8, %xmm7
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm9, %xmm0
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm0, %xmm0
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm12[3,4,5],xmm0[6,7]
-; AVX1-ONLY-NEXT: vandnps %ymm11, %ymm6, %ymm7
-; AVX1-ONLY-NEXT: vandps %ymm6, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vorps %ymm7, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm6, %xmm8
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm5, %xmm1
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = xmm4[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm4[5,11]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm10 = xmm2[u,u,u,u,u,u,u,u,u,u,u,3,9,15],zero,zero
+; AVX1-ONLY-NEXT: vpor %xmm8, %xmm10, %xmm8
+; AVX1-ONLY-NEXT: vpblendvb %xmm3, %xmm1, %xmm8, %xmm1
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm11, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm11, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,zero,xmm5[4,10,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm6 = xmm3[2,8,14],zero,zero,xmm3[u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm6, %xmm1
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm6 = xmm4[u,u,u,u,u,u,u,u,u,u,u,4,10],zero,zero,zero
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = xmm2[u,u,u,u,u,u,u,u,u,u,u],zero,zero,xmm2[0,6,12]
-; AVX1-ONLY-NEXT: vpor %xmm6, %xmm7, %xmm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm6, %ymm1
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm6 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255]
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm6, %ymm1
-; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm7, %ymm7
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = zero,zero,zero,xmm9[4,10,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm13[2,8,14],zero,zero,xmm13[u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [0,0,0,4,10,128,128,128,0,0,0,4,10,128,128,128]
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm7, %xmm10
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [0,0,0,128,128,0,6,12,0,0,0,128,128,0,6,12]
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm12, %xmm14
+; AVX1-ONLY-NEXT: vpor %xmm10, %xmm14, %xmm10
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm10, %ymm0
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm10, %ymm0
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm10 = xmm5[2,8,14,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm14 = xmm6[u,u,u,u,u,0,6,12,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm10 = xmm14[0],xmm10[0]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm2, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm4, %xmm11
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm11, %xmm1
+; AVX1-ONLY-NEXT: vpblendvb %xmm3, %xmm10, %xmm1, %xmm1
; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm10 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm10, %ymm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm8, %ymm10, %ymm8
-; AVX1-ONLY-NEXT: vorps %ymm7, %ymm8, %ymm13
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,zero,xmm5[5,11,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = xmm3[3,9,15],zero,zero,xmm3[u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm8, %xmm7
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = xmm4[u,u,u,u,u,u,u,u,u,u,u,5,11],zero,zero,zero
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm9 = xmm2[u,u,u,u,u,u,u,u,u,u,u],zero,zero,xmm2[1,7,13]
-; AVX1-ONLY-NEXT: vpor %xmm8, %xmm9, %xmm8
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm8, %ymm7
-; AVX1-ONLY-NEXT: vandnps %ymm7, %ymm6, %ymm7
-; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vorps %ymm7, %ymm6, %ymm6
-; AVX1-ONLY-NEXT: vandps %ymm6, %ymm10, %ymm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm7, %ymm10, %ymm7
-; AVX1-ONLY-NEXT: vorps %ymm7, %ymm6, %ymm14
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm6 = xmm3[4,10],zero,zero,zero,xmm3[u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,xmm5[0,6,12,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm6, %xmm7, %xmm6
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [0,0,128,128,128,2,8,14,0,0,128,128,128,2,8,14]
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm2, %xmm8
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = [0,0,0,6,12,128,128,128,0,0,0,6,12,128,128,128]
-; AVX1-ONLY-NEXT: # xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm4, %xmm11
-; AVX1-ONLY-NEXT: vpor %xmm8, %xmm11, %xmm8
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm8, %ymm6
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm10, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm10, %ymm1
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm11
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = zero,zero,zero,xmm9[5,11,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm13[3,9,15],zero,zero,xmm13[u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [0,0,0,5,11,128,128,128,0,0,0,5,11,128,128,128]
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm7, %xmm14
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm15 = [0,0,0,128,128,1,7,13,0,0,0,128,128,1,7,13]
+; AVX1-ONLY-NEXT: # xmm15 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm12, %xmm8
+; AVX1-ONLY-NEXT: vpor %xmm14, %xmm8, %xmm8
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm8, %ymm0
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = xmm5[3,9,15,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm14 = xmm6[u,u,u,u,u,1,7,13,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm8 = xmm14[0],xmm8[0]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm2, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm4, %xmm14
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm14, %xmm1
+; AVX1-ONLY-NEXT: vpblendvb %xmm3, %xmm8, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm10, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm10, %ymm1
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm13[4,10],zero,zero,zero,xmm13[u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,xmm9[0,6,12,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [0,0,128,128,128,2,8,14,0,0,128,128,128,2,8,14]
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm12, %xmm8
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm14 = [0,0,0,6,12,128,128,128,0,0,0,6,12,128,128,128]
+; AVX1-ONLY-NEXT: # xmm14 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm7, %xmm15
+; AVX1-ONLY-NEXT: vpor %xmm8, %xmm15, %xmm8
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm8, %ymm0
; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm8 = [0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vandps %ymm6, %ymm8, %ymm6
-; AVX1-ONLY-NEXT: vorps %ymm6, %ymm11, %ymm6
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm1, %xmm7
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm0, %xmm9
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm9, %xmm7
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm9 = xmm15[u,u,u,u,u,u,u,u,4,10,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm12[u,u,u,u,u,u,u,u,u,u,u,u,u,2,8,14]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm9 = xmm11[1],xmm9[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm9[0,1,2,3,4],xmm7[5,6,7]
-; AVX1-ONLY-NEXT: vandps %ymm6, %ymm10, %ymm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vandnps %ymm7, %ymm10, %ymm7
-; AVX1-ONLY-NEXT: vorps %ymm7, %ymm6, %ymm6
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[5,11],zero,zero,zero,xmm3[u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,xmm5[1,7,13,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm5, %xmm3
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = [0,0,128,128,128,3,9,15,0,0,128,128,128,3,9,15]
-; AVX1-ONLY-NEXT: # xmm5 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [0,0,1,7,13,128,128,128,0,0,1,7,13,128,128,128]
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm4, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm4, %xmm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vandps %ymm2, %ymm8, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm1, %xmm3
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm0, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm4, %xmm3
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm4 = xmm15[u,u,u,u,u,u,u,u,5,11,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm5 = xmm12[u,u,u,u,u,u,u,u,u,u,u,u,u,3,9,15]
+; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm8, %ymm0
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm15, %ymm0
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm4, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm2, %xmm14
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm14, %xmm1
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm14 = xmm5[u,u,u,u,u,u,u,u,4,10,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm15 = xmm6[u,u,u,u,u,u,u,u,u,u,u,u,u,2,8,14]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm14 = xmm15[1],xmm14[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm14[0,1,2,3,4],xmm1[5,6,7]
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm10, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm10, %ymm1
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm13[5,11],zero,zero,zero,xmm13[u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm9 = zero,zero,xmm9[1,7,13,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm9, %xmm1
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = [0,0,128,128,128,3,9,15,0,0,128,128,128,3,9,15]
+; AVX1-ONLY-NEXT: # xmm9 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm12, %xmm12
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm13 = [0,0,1,7,13,128,128,128,0,0,1,7,13,128,128,128]
+; AVX1-ONLY-NEXT: # xmm13 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm7, %xmm7
+; AVX1-ONLY-NEXT: vpor %xmm7, %xmm12, %xmm7
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm7, %ymm1
+; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm8, %ymm1
+; AVX1-ONLY-NEXT: vorps %ymm7, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm4, %xmm4
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm4 = xmm5[u,u,u,u,u,u,u,u,5,11,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm5 = xmm6[u,u,u,u,u,u,u,u,u,u,u,u,u,3,9,15]
; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm4 = xmm5[1],xmm4[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX1-ONLY-NEXT: vandps %ymm2, %ymm10, %ymm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm10, %ymm1
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm13, (%rcx)
-; AVX1-ONLY-NEXT: vmovaps %ymm14, (%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm6, (%r9)
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2,3,4],xmm2[5,6,7]
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm10, %ymm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm11, (%rcx)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, (%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rax)
-; AVX1-ONLY-NEXT: addq $168, %rsp
+; AVX1-ONLY-NEXT: addq $120, %rsp
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -3586,646 +3588,627 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
;
; AVX1-ONLY-LABEL: load_i8_stride6_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $808, %rsp # imm = 0x328
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = [0,0,0,128,128,128,4,10,0,0,0,128,128,128,4,10]
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [0,0,0,2,8,14,128,128,0,0,0,2,8,14,128,128]
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm2 = [0,0,4,10,0,0,4,10,0,0,4,10,0,0,4,10]
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [0,6,12,0,0,6,12,0,0,6,12,0,0,6,12,0]
-; AVX1-ONLY-NEXT: vmovdqa 128(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm1, %xmm0
+; AVX1-ONLY-NEXT: subq $616, %rsp # imm = 0x268
+; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vmovdqa 32(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovdqa 48(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovdqa 224(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovdqa 240(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovdqa 208(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [128,128,128,4,10,0,0,0,128,128,128,4,10,0,0,0]
+; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm12 = [2,8,14,128,128,0,0,0,2,8,14,128,128,0,0,0]
+; AVX1-ONLY-NEXT: # xmm12 = mem[0,0]
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm1 = [2,8,14,0,2,8,14,0,2,8,14,0,2,8,14,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm6, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm13
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm4 = [0,0,6,12,0,0,6,12,0,0,6,12,0,0,6,12]
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm5, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm15
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm14
+; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm7, %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm8
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 144(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm5
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm10
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm2, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm9
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 160(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm11
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm2, %xmm2
; AVX1-ONLY-NEXT: vpor %xmm0, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,u,u,255,255,255,255,255,0,0,0,0,0>
-; AVX1-ONLY-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm12
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 336(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm13
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm14
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm4, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm15
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm2 = [0,0,5,11,0,0,5,11,0,0,5,11,0,0,5,11]
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [1,7,13,0,1,7,13,0,1,7,13,0,1,7,13,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm8, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm4
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm5 = <0,0,0,0,0,255,255,255,255,255,255,u,u,u,u,u>
+; AVX1-ONLY-NEXT: vpblendvb %xmm5, %xmm1, %xmm2, %xmm1
+; AVX1-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm9, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm8, %xmm2
+; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm10, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm5
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [0,0,0,128,128,128,5,11,0,0,0,128,128,128,5,11]
+; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm11, %xmm3
+; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
+; AVX1-ONLY-NEXT: vpblendvb %xmm5, %xmm1, %xmm2, %xmm1
+; AVX1-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm2 = [3,9,15,0,3,9,15,0,3,9,15,0,3,9,15,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm6, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm13
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm0 = [0,1,7,13,0,1,7,13,0,1,7,13,0,1,7,13]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm15, %xmm2
+; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [128,128,128,5,11,0,0,0,128,128,128,5,11,0,0,0]
; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = [0,0,0,3,9,15,128,128,0,0,0,3,9,15,128,128]
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm9, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm7
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm11, %xmm3
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm12 = [3,9,15,128,128,0,0,0,3,9,15,128,128,0,0,0]
+; AVX1-ONLY-NEXT: # xmm12 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm7, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm7, %xmm14
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm15
+; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm4, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm7
; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm12, %xmm1
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm13, %xmm2
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm14, %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm15, %xmm3
+; AVX1-ONLY-NEXT: vpblendvb %xmm5, %xmm1, %xmm2, %xmm1
+; AVX1-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm9, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm8, %xmm2
+; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm10, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm11, %xmm3
; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 32(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovdqa 48(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm0 = [2,8,14,0,2,8,14,0,2,8,14,0,2,8,14,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm11, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, %xmm8
-; AVX1-ONLY-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm15 = [0,0,6,12,0,0,6,12,0,0,6,12,0,0,6,12]
-; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm4, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm10
-; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [128,128,128,4,10,0,0,0,128,128,128,4,10,0,0,0]
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = [2,8,14,128,128,0,0,0,2,8,14,128,128,0,0,0]
-; AVX1-ONLY-NEXT: # xmm5 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm14, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm12
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm10, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm6
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm4, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm13 = <0,0,0,0,0,255,255,255,255,255,255,u,u,u,u,u>
-; AVX1-ONLY-NEXT: vpblendvb %xmm13, %xmm3, %xmm4, %xmm3
-; AVX1-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 224(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovdqa 240(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm9, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm5, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm8
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm3 = xmm4[0],xmm3[0]
-; AVX1-ONLY-NEXT: vmovdqa 208(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm1, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm2, %xmm5
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpor %xmm4, %xmm5, %xmm4
-; AVX1-ONLY-NEXT: vpblendvb %xmm13, %xmm3, %xmm4, %xmm3
-; AVX1-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [3,9,15,0,3,9,15,0,3,9,15,0,3,9,15,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm11, %xmm5
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm11
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm12 = [0,1,7,13,0,1,7,13,0,1,7,13,0,1,7,13]
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm0, %xmm6
-; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm5 = xmm6[0],xmm5[0]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [128,128,128,5,11,0,0,0,128,128,128,5,11,0,0,0]
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [3,9,15,128,128,0,0,0,3,9,15,128,128,0,0,0]
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm14, %xmm6
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm10, %xmm7
-; AVX1-ONLY-NEXT: vpor %xmm6, %xmm7, %xmm6
-; AVX1-ONLY-NEXT: vpblendvb %xmm13, %xmm5, %xmm6, %xmm0
-; AVX1-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm9, %xmm5
-; AVX1-ONLY-NEXT: vmovdqa %xmm11, %xmm14
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm8, %xmm6
-; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm5 = xmm6[0],xmm5[0]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm1, %xmm6
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm2, %xmm7
-; AVX1-ONLY-NEXT: vpor %xmm6, %xmm7, %xmm6
-; AVX1-ONLY-NEXT: vpblendvb %xmm13, %xmm5, %xmm6, %xmm0
+; AVX1-ONLY-NEXT: vpblendvb %xmm5, %xmm1, %xmm2, %xmm0
; AVX1-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm1 = [2,8,14,0,2,8,14,0,2,8,14,0,2,8,14,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm9, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm8, %xmm5
-; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm5[0],xmm2[0]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [0,0,0,4,10,128,128,128,0,0,0,4,10,128,128,128]
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [0,0,0,128,128,0,6,12,0,0,0,128,128,0,6,12]
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm11, %xmm5
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm10, %xmm6
-; AVX1-ONLY-NEXT: vpor %xmm5, %xmm6, %xmm5
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm6 = <u,u,u,u,u,255,255,255,255,255,255,0,0,0,0,0>
-; AVX1-ONLY-NEXT: vpblendvb %xmm6, %xmm2, %xmm5, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm13, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm5, %xmm1
-; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm15, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm7, %xmm2
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vpblendvb %xmm6, %xmm0, %xmm1, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm9, %xmm0
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm8, %xmm1
-; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [0,0,0,5,11,128,128,128,0,0,0,5,11,128,128,128]
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm8 = [0,0,0,128,128,1,7,13,0,0,0,128,128,1,7,13]
-; AVX1-ONLY-NEXT: # xmm8 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm11, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm3
-; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm10, %xmm2
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vpblendvb %xmm6, %xmm0, %xmm1, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm13, %xmm0
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm5, %xmm1
-; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm15, %xmm1
-; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm7, %xmm2
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vpblendvb %xmm6, %xmm0, %xmm1, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [4,10,128,128,128,0,0,0,4,10,128,128,128,0,0,0]
; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm10, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm5
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm4, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm12
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [128,128,0,6,12,0,0,0,128,128,0,6,12,0,0,0]
; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm4, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm9
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm14, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm13
; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm1
; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm2 = [4,10,0,0,4,10,0,0,4,10,0,0,4,10,0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm12, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm6
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm6, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm4
; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [0,2,8,14,0,2,8,14,0,2,8,14,0,2,8,14]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm14, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm5, %xmm2
; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = <255,255,255,255,255,0,0,0,0,0,u,u,u,u,u,u>
; AVX1-ONLY-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm1
; AVX1-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm15, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm8, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm11, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm10, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm10, %xmm12
; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm7, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm6, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm9, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm9, %xmm15
+; AVX1-ONLY-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm8, %xmm3
; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm1
; AVX1-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [5,11,128,128,128,0,0,0,5,11,128,128,128,0,0,0]
; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm10, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm9
+; AVX1-ONLY-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm7, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm8
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [128,128,1,7,13,0,0,0,128,128,1,7,13,0,0,0]
; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm4, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm10
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm13
+; AVX1-ONLY-NEXT: vmovdqa %xmm14, %xmm13
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm14, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm14
; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [5,11,0,0,5,11,0,0,5,11,0,0,5,11,0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm12, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm11
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm6, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm9
+; AVX1-ONLY-NEXT: vmovdqa %xmm6, %xmm10
+; AVX1-ONLY-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm4 = [0,3,9,15,0,3,9,15,0,3,9,15,0,3,9,15]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm14, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm5, %xmm3
; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm1
; AVX1-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm15, %xmm1
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm8, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm11, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm11, %xmm8
+; AVX1-ONLY-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm12, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm7, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm15, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm6, %xmm3
; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm0
; AVX1-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [128,128,128,2,8,14,0,0,128,128,128,2,8,14,0,0]
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [128,128,128,2,8,14,0,0,128,128,128,2,8,14,0,0]
+; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
; AVX1-ONLY-NEXT: vmovdqa 112(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm3
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [0,6,12,128,128,128,0,0,0,6,12,128,128,128,0,0]
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm15 = [0,6,12,128,128,128,0,0,0,6,12,128,128,128,0,0]
+; AVX1-ONLY-NEXT: # xmm15 = mem[0,0]
; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm5
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm1, %xmm1
; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa 80(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm10 = [0,0,0,128,128,128,4,10,0,0,0,128,128,128,4,10]
-; AVX1-ONLY-NEXT: # xmm10 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm9
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [0,0,0,128,128,128,4,10,0,0,0,128,128,128,4,10]
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa 80(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm11
; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [0,0,0,2,8,14,128,128,0,0,0,2,8,14,128,128]
-; AVX1-ONLY-NEXT: # xmm11 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm2, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm4, %xmm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm14[u,u,4,10,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm4 = xmm12[u,u,u,u,0,6,12,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm13, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm14
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm12
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm2, %xmm5
-; AVX1-ONLY-NEXT: vpor %xmm4, %xmm5, %xmm4
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2],xmm0[3,4,5],xmm4[6,7]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,u,u,u,u,u,2,8,14],zero,zero
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm2
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm14 = [0,0,4,10,0,0,4,10,0,0,4,10,0,0,4,10]
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm5, %xmm4
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm1 = [0,6,12,0,0,6,12,0,0,6,12,0,0,6,12,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm10, %xmm5
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm13, %xmm5
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm7, %xmm9
+; AVX1-ONLY-NEXT: vpor %xmm5, %xmm9, %xmm5
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3,4,5],xmm5[6,7]
; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm13 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255]
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm13, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm13, %ymm2
; AVX1-ONLY-NEXT: vandps %ymm4, %ymm13, %ymm4
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm3 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandps %ymm3, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm3, %ymm5
-; AVX1-ONLY-NEXT: vorps %ymm5, %ymm4, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 304(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm0, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm0, %xmm5
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, %xmm12
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpor %xmm4, %xmm5, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa 272(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm0, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa 256(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm4, %ymm9
+; AVX1-ONLY-NEXT: vmovdqa 128(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 144(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm4, %xmm4
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 160(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [0,0,0,2,8,14,128,128,0,0,0,2,8,14,128,128]
+; AVX1-ONLY-NEXT: # xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm5, %xmm10
+; AVX1-ONLY-NEXT: vpor %xmm2, %xmm10, %xmm10
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255]
+; AVX1-ONLY-NEXT: vpblendvb %xmm2, %xmm4, %xmm10, %xmm10
+; AVX1-ONLY-NEXT: vmovdqa %ymm2, %ymm5
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm2 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm9, %ymm9
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vandnps %ymm10, %ymm2, %ymm10
+; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm7
+; AVX1-ONLY-NEXT: vorps %ymm10, %ymm9, %ymm4
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 304(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm2, %xmm9
+; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm2, %xmm10
+; AVX1-ONLY-NEXT: vpor %xmm9, %xmm10, %xmm9
+; AVX1-ONLY-NEXT: vmovdqa 272(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm2, %xmm10
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm10, %xmm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm6[u,u,4,10,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm10 = xmm7[u,u,u,u,0,6,12,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
-; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm8, %xmm7
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm6 = xmm15[0,6,12],zero,zero,zero,xmm15[u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm6[0,1,2],xmm1[3,4,5],xmm6[6,7]
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm13, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm13, %ymm1
+; AVX1-ONLY-NEXT: vmovdqa 256(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm2, %xmm11
+; AVX1-ONLY-NEXT: vpor %xmm10, %xmm11, %xmm10
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm10, %ymm9
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm6, %xmm10
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm2, %xmm11
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm12, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm8, %xmm6
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm6, %xmm3
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm10[3,4,5],xmm3[6,7]
+; AVX1-ONLY-NEXT: vandnps %ymm9, %ymm13, %ymm6
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm13, %ymm3
+; AVX1-ONLY-NEXT: vmovaps %ymm13, %ymm11
+; AVX1-ONLY-NEXT: vorps %ymm6, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm4, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa 336(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm4, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm1[4,10]
+; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm6 = xmm4[u,u,u,u,u,u,u,u,u,u,u,2,8,14],zero,zero
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm6, %xmm1
+; AVX1-ONLY-NEXT: vpblendvb %xmm5, %xmm0, %xmm1, %xmm0
+; AVX1-ONLY-NEXT: vandps %ymm7, %ymm3, %ymm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm7, %ymm0
; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm3, %ymm1
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [128,128,128,3,9,15,0,0,128,128,128,3,9,15,0,0]
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm14, %xmm1
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = [1,7,13,128,128,128,0,0,1,7,13,128,128,128,0,0]
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm15, %xmm7
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm7, %xmm1
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm8 = [0,0,0,128,128,128,5,11,0,0,0,128,128,128,5,11]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm8 = [128,128,128,3,9,15,0,0,128,128,128,3,9,15,0,0]
; AVX1-ONLY-NEXT: # xmm8 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm9, %xmm7
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [1,7,13,128,128,128,0,0,1,7,13,128,128,128,0,0]
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm1[5,11]
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = [0,0,0,3,9,15,128,128,0,0,0,3,9,15,128,128]
; AVX1-ONLY-NEXT: # xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm2, %xmm10
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm10, %xmm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm7, %ymm1
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm4 = [0,0,5,11,0,0,5,11,0,0,5,11,0,0,5,11]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm5, %xmm7
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm5 = [1,7,13,0,1,7,13,0,1,7,13,0,1,7,13,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm10, %xmm10
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm10[0],xmm7[0],xmm10[1],xmm7[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm10, %xmm10
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm11, %xmm11
-; AVX1-ONLY-NEXT: vpor %xmm10, %xmm11, %xmm10
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm10[0,1,2],xmm7[3,4,5],xmm10[6,7]
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm13, %ymm1
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm13, %ymm7
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm7, %ymm1
-; AVX1-ONLY-NEXT: vandps %ymm3, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm7, %ymm3, %ymm7
-; AVX1-ONLY-NEXT: vorps %ymm7, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm11, %xmm1
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm12, %xmm7
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm7, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm5, %xmm14
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm14, %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm14
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [0,0,5,11,0,0,5,11,0,0,5,11,0,0,5,11]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm7 = [1,7,13,0,1,7,13,0,1,7,13,0,1,7,13,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm6, %xmm15
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm6, %xmm15
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm6, %xmm13
+; AVX1-ONLY-NEXT: vpor %xmm15, %xmm13, %xmm13
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm13[0,1,2],xmm0[3,4,5],xmm13[6,7]
+; AVX1-ONLY-NEXT: vandnps %ymm14, %ymm11, %ymm13
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm11, %ymm0
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm13, %ymm0
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm12, %xmm7
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm8, %xmm10
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm10, %xmm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm7, %ymm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm7, %xmm7
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm4, %xmm10
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm10[0],xmm7[0],xmm10[1],xmm7[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm4, %xmm0
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm12, %xmm13
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm11, %xmm14
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [0,0,0,128,128,128,5,11,0,0,0,128,128,128,5,11]
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm6, %xmm14
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm15, %xmm15
+; AVX1-ONLY-NEXT: vpor %xmm14, %xmm15, %xmm14
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} ymm15 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255]
+; AVX1-ONLY-NEXT: vpblendvb %xmm15, %xmm13, %xmm14, %xmm13
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm15 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm15, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm15, %ymm13
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm13, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm13, %xmm13
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm13, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm13, %xmm13
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm14, %xmm14
+; AVX1-ONLY-NEXT: vpor %xmm13, %xmm14, %xmm13
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm13, %ymm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm13, %xmm13
+; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm2, %xmm14
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm2, %xmm8
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm2, %xmm4
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm8, %xmm4
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2],xmm13[3,4,5],xmm4[6,7]
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255]
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm4, %ymm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm2, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm10, %xmm1
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm14[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm14[5,11]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm4, %xmm6
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm6, %xmm0
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm7[3,4,5],xmm0[6,7]
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm13, %ymm1
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm13, %ymm0
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm3, %ymm1
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm4, %xmm4
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm4, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} ymm8 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255]
+; AVX1-ONLY-NEXT: vpblendvb %xmm8, %xmm1, %xmm3, %xmm1
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm15, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm15, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [128,128,128,4,10,0,0,0,128,128,128,4,10,0,0,0]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = [128,128,128,4,10,0,0,0,128,128,128,4,10,0,0,0]
+; AVX1-ONLY-NEXT: # xmm9 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm10, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm13 = [2,8,14,128,128,0,0,0,2,8,14,128,128,0,0,0]
+; AVX1-ONLY-NEXT: # xmm13 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [0,0,0,4,10,128,128,128,0,0,0,4,10,128,128,128]
+; AVX1-ONLY-NEXT: # xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm5, %xmm4
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [0,0,0,128,128,0,6,12,0,0,0,128,128,0,6,12]
; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa %xmm14, %xmm13
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm14, %xmm0
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [2,8,14,128,128,0,0,0,2,8,14,128,128,0,0,0]
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm15, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm15, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm0
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm15 = [0,0,0,4,10,128,128,128,0,0,0,4,10,128,128,128]
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm5
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = [0,0,0,128,128,0,6,12,0,0,0,128,128,0,6,12]
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm5, %xmm5
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm5, %xmm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm4, %ymm1
+; AVX1-ONLY-NEXT: vmovdqa %ymm8, %ymm9
+; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm8, %ymm1
+; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm13 = [2,8,14,0,2,8,14,0,2,8,14,0,2,8,14,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm11, %xmm1
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm8 = [0,0,6,12,0,0,6,12,0,0,6,12,0,0,6,12]
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm12, %xmm5
+; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm5[0],xmm1[0]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm7, %xmm2
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255]
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm14 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm14, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm14, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa %xmm11, %xmm9
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm11, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm10[2,8,14],zero,zero,xmm10[u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm8, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm8, %xmm15
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm12, %xmm6
-; AVX1-ONLY-NEXT: vmovdqa %xmm12, %xmm11
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm6, %xmm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm2, %ymm14, %ymm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm14, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm1
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm7, %xmm5
+; AVX1-ONLY-NEXT: vmovdqa %xmm6, %xmm15
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm6, %xmm6
+; AVX1-ONLY-NEXT: vpor %xmm5, %xmm6, %xmm5
+; AVX1-ONLY-NEXT: vpblendvb %xmm9, %xmm1, %xmm5, %xmm5
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm1, %ymm5
+; AVX1-ONLY-NEXT: vorps %ymm5, %ymm4, %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [128,128,128,5,11,0,0,0,128,128,128,5,11,0,0,0]
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm13, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm13, %xmm1
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [3,9,15,128,128,0,0,0,3,9,15,128,128,0,0,0]
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm13
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm4, %xmm6
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm6, %xmm3
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [0,0,0,5,11,128,128,128,0,0,0,5,11,128,128,128]
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm12
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm5, %xmm6
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm4 = zero,zero,zero,xmm12[4,10,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm5 = xmm1[2,8,14],zero,zero,xmm1[u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm5, %xmm4
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm11, %xmm5
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm6, %xmm6
+; AVX1-ONLY-NEXT: vpor %xmm5, %xmm6, %xmm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm5, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm9, %ymm4
+; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vorps %ymm4, %ymm5, %ymm4
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm1, %xmm5
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm2, %xmm6
+; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm5 = xmm6[0],xmm5[0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm2, %xmm0
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm14, %xmm3
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm3, %xmm0
+; AVX1-ONLY-NEXT: vpblendvb %xmm9, %xmm5, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm4, %ymm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm5, %ymm0
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm14 = [128,128,128,5,11,0,0,0,128,128,128,5,11,0,0,0]
+; AVX1-ONLY-NEXT: # xmm14 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm10, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm13 = [3,9,15,128,128,0,0,0,3,9,15,128,128,0,0,0]
+; AVX1-ONLY-NEXT: # xmm13 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm14, %xmm3
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm3, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [0,0,0,5,11,128,128,128,0,0,0,5,11,128,128,128]
+; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm4, %xmm4
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = [0,0,0,128,128,1,7,13,0,0,0,128,128,1,7,13]
; AVX1-ONLY-NEXT: # xmm5 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa %xmm7, %xmm8
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm7, %xmm7
-; AVX1-ONLY-NEXT: vpor %xmm6, %xmm7, %xmm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm3
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm6, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm3, %ymm14, %ymm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm14, %ymm6
-; AVX1-ONLY-NEXT: vorps %ymm6, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,zero,xmm9[5,11,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm10, %xmm7
-; AVX1-ONLY-NEXT: vmovdqa %xmm10, %xmm9
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm10, %xmm6
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm6, %xmm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm4, %ymm0
+; AVX1-ONLY-NEXT: vmovdqa %ymm9, %ymm13
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm9, %ymm0
+; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm4, %ymm0
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm8 = [3,9,15,0,3,9,15,0,3,9,15,0,3,9,15,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm4, %xmm4
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm9 = [0,1,7,13,0,1,7,13,0,1,7,13,0,1,7,13]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm6, %xmm6
+; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm4 = xmm6[0],xmm4[0]
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm7, %xmm6
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm15, %xmm7
+; AVX1-ONLY-NEXT: vmovdqa %xmm15, %xmm8
; AVX1-ONLY-NEXT: vpor %xmm6, %xmm7, %xmm6
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm15, %xmm7
-; AVX1-ONLY-NEXT: vmovdqa %xmm15, %xmm4
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm11, %xmm10
-; AVX1-ONLY-NEXT: vmovdqa %xmm11, %xmm2
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm10, %xmm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm7, %ymm6
-; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vorps %ymm6, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm14, %ymm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm14, %ymm6
-; AVX1-ONLY-NEXT: vmovaps %ymm14, %ymm11
-; AVX1-ONLY-NEXT: vorps %ymm6, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vpblendvb %xmm13, %xmm4, %xmm6, %xmm4
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm6 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandps %ymm6, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm6, %ymm4
+; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vorps %ymm4, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm14 = [4,10,128,128,128,0,0,0,4,10,128,128,128,0,0,0]
-; AVX1-ONLY-NEXT: # xmm14 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm13, %xmm0
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm15 = [128,128,0,6,12,0,0,0,128,128,0,6,12,0,0,0]
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm1, %xmm7
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm7, %xmm0
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [0,0,128,128,128,2,8,14,0,0,128,128,128,2,8,14]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = zero,zero,zero,xmm12[5,11,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm4 = xmm12[3,9,15],zero,zero,xmm12[u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm4, %xmm0
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm11, %xmm4
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm15, %xmm6
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm6, %xmm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm4, %ymm0
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm4 = xmm1[3,9,15,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm1, %xmm6
+; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm4 = xmm6[0],xmm4[0]
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm2, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm11, %xmm5
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm5, %xmm3
+; AVX1-ONLY-NEXT: vpblendvb %xmm13, %xmm4, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm13, %ymm0
+; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm2, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm7, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm7, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [4,10,128,128,128,0,0,0,4,10,128,128,128,0,0,0]
; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm8, %xmm7
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = [0,0,0,6,12,128,128,128,0,0,0,6,12,128,128,128]
-; AVX1-ONLY-NEXT: # xmm5 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm12, %xmm10
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm10, %xmm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm7, %ymm0
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm14, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm14 = [128,128,0,6,12,0,0,0,128,128,0,6,12,0,0,0]
+; AVX1-ONLY-NEXT: # xmm14 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm13, %xmm2
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [0,0,128,128,128,2,8,14,0,0,128,128,128,2,8,14]
+; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm10, %xmm2
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [0,0,0,6,12,128,128,128,0,0,0,6,12,128,128,128]
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm5, %xmm5
+; AVX1-ONLY-NEXT: vpor %xmm2, %xmm5, %xmm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm10 = [0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm10, %ymm0
-; AVX1-ONLY-NEXT: vorps %ymm7, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm15, %xmm7
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm8, %xmm3
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm8, %xmm5
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm0, %xmm6
+; AVX1-ONLY-NEXT: vpor %xmm5, %xmm6, %xmm5
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm8 = [4,10,0,0,4,10,0,0,4,10,0,0,4,10,0,0]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm12 = [4,10,0,0,4,10,0,0,4,10,0,0,4,10,0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm14, %xmm7
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm14, %xmm6
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm9 = [0,2,8,14,0,2,8,14,0,2,8,14,0,2,8,14]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm0, %xmm7
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2,3,4],xmm5[5,6,7]
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm6 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandps %ymm6, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm6, %ymm5
+; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm0
+; AVX1-ONLY-NEXT: vorps %ymm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm12, %xmm5
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,xmm1[0,6,12,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm5, %xmm6, %xmm5
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm15, %xmm6
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm1, %xmm7
+; AVX1-ONLY-NEXT: vpor %xmm6, %xmm7, %xmm6
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm6, %ymm5
+; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm10, %ymm5
+; AVX1-ONLY-NEXT: vorps %ymm6, %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm11, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm11, %xmm4
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm4, %xmm3
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm13 = [0,2,8,14,0,2,8,14,0,2,8,14,0,2,8,14]
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm12, %xmm6
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm7[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm6[0,1,2,3,4],xmm3[5,6,7]
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm11, %ymm0
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm12, %xmm4
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm1, %xmm6
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm4 = xmm6[1],xmm4[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm5, %ymm4
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm11, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm9[4,10],zero,zero,zero,xmm9[u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,xmm7[0,6,12,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm3, %xmm0
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm2, %xmm3
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm4, %xmm6
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm6, %xmm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
-; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm10, %ymm0
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm11, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm9, %xmm1
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm3, %ymm4, %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm15 = [5,11,128,128,128,0,0,0,5,11,128,128,128,0,0,0]
+; AVX1-ONLY-NEXT: # xmm15 = mem[0,0]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,u,u,4,10,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm4, %xmm3
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm3[1],xmm2[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4],xmm1[5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm2, %xmm3
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [128,128,1,7,13,0,0,0,128,128,1,7,13,0,0,0]
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm13, %xmm5
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm5, %xmm3
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = [0,0,128,128,128,3,9,15,0,0,128,128,128,3,9,15]
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm2, %xmm7
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = [0,0,1,7,13,128,128,128,0,0,1,7,13,128,128,128]
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm2, %xmm8
+; AVX1-ONLY-NEXT: vpor %xmm7, %xmm8, %xmm7
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm7, %ymm3
+; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm10, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm7, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm2, %xmm7
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm0, %xmm8
+; AVX1-ONLY-NEXT: vpor %xmm7, %xmm8, %xmm7
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm13 = [5,11,0,0,5,11,0,0,5,11,0,0,5,11,0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm14, %xmm8
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm14 = [0,3,9,15,0,3,9,15,0,3,9,15,0,3,9,15]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm0, %xmm9
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm8 = xmm9[1],xmm8[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm8[0,1,2,3,4],xmm7[5,6,7]
; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT: vandnps %ymm7, %ymm2, %ymm7
+; AVX1-ONLY-NEXT: vorps %ymm7, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm7, %xmm7
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm8, %xmm8
+; AVX1-ONLY-NEXT: vpor %xmm7, %xmm8, %xmm7
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm4, %xmm8
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm4, %xmm9
+; AVX1-ONLY-NEXT: vpor %xmm8, %xmm9, %xmm8
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm8, %ymm7
+; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vandps %ymm7, %ymm10, %ymm0
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm8, %ymm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm4, %xmm5
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm11, %xmm6
+; AVX1-ONLY-NEXT: vpor %xmm5, %xmm6, %xmm5
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm12, %xmm6
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm1, %xmm7
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2,3,4],xmm5[5,6,7]
; AVX1-ONLY-NEXT: vandps %ymm2, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm2, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm0[5,11],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,xmm0[1,7,13,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm13 = [0,0,128,128,128,3,9,15,0,0,128,128,128,3,9,15]
-; AVX1-ONLY-NEXT: # xmm13 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm0, %xmm2
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [0,0,1,7,13,128,128,128,0,0,1,7,13,128,128,128]
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm3, %xmm3
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm10, %ymm1
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm15, %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm8, %xmm3
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm8 = [5,11,0,0,5,11,0,0,5,11,0,0,5,11,0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm14, %xmm3
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm5 = [0,3,9,15,0,3,9,15,0,3,9,15,0,3,9,15]
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm12, %xmm6
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm3 = xmm6[1],xmm3[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3,4],xmm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandps %ymm3, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm3, %ymm2
-; AVX1-ONLY-NEXT: vmovaps %ymm3, %ymm12
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm1, %ymm6
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[5,11],zero,zero,zero,xmm1[u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,xmm7[1,7,13,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm3, %xmm3
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm10, %ymm1
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm11, %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm9, %xmm3
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm0, %xmm3
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm4, %xmm4
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm3 = xmm4[1],xmm3[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3,4],xmm2[5,6,7]
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm12, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm12, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm6, (%rax)
-; AVX1-ONLY-NEXT: addq $808, %rsp # imm = 0x328
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rax)
+; AVX1-ONLY-NEXT: addq $616, %rsp # imm = 0x268
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
index 0e3f8121db1de0..f82a192c60b570 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
@@ -9218,515 +9218,508 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
;
; AVX512F-ONLY-SLOW-LABEL: load_i8_stride7_vf64:
; AVX512F-ONLY-SLOW: # %bb.0:
-; AVX512F-ONLY-SLOW-NEXT: subq $72, %rsp
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = [65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 (%rdi), %ymm17
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 32(%rdi), %ymm15
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 64(%rdi), %ymm24
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm1, %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm1, %ymm12
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm15, %ymm17, %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,zero,xmm1[5,12],zero,zero,xmm1[1,8,15,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,7,14],zero,zero,xmm0[3,10],zero,zero,zero,xmm0[u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm1, %xmm0, %xmm1
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm11 = [65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm0 = [65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa (%rdi), %ymm12
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 32(%rdi), %ymm13
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 64(%rdi), %ymm31
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm0, %ymm1
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm0, %ymm24
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm13, %ymm12, %ymm1
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm2[5,12],zero,zero,xmm2[1,8,15,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,7,14],zero,zero,xmm1[3,10],zero,zero,zero,xmm1[u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm2, %xmm1, %xmm1
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm9 = [65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535]
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 96(%rdi), %ymm19
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm11, %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm24, %ymm19, %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 80(%rdi), %xmm0
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm0[2],ymm2[3,4],ymm0[5],ymm2[6,7,8,9],ymm0[10],ymm2[11,12],ymm0[13],ymm2[14,15]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm31, %ymm19, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 80(%rdi), %xmm11
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm11[2],ymm2[3,4],ymm11[5],ymm2[6,7,8,9],ymm11[10],ymm2[11,12],ymm11[13],ymm2[14,15]
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[6,13,4,11,2,9,16,23,30,u],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 128(%rdi), %ymm31
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm14 = [65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 128(%rdi), %ymm21
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 160(%rdi), %ymm29
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm1, %ymm4
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm31, %ymm29, %ymm4
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u],zero,zero,xmm5[3,10],zero,zero,zero,xmm5[6,13,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,5,12],zero,zero,xmm4[1,8,15],zero,zero,xmm4[u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm4, %xmm4
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 192(%rdi), %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm5[u,u,u,u,u,u,4,11,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm5, %xmm6
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 208(%rdi), %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm9 = xmm5[u,u,u,u,u,u,2,9,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm5, %xmm23
-; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1],xmm7[2],xmm9[2],xmm7[3],xmm9[3]
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm7[7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 240(%rdi), %xmm14
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,zero,xmm14[5,12,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa 224(%rdi), %xmm8
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm9 = xmm8[0,7,14],zero,zero,xmm8[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm9, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm7, %zmm4, %zmm7
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm26 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm2, %zmm26, %zmm7
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 288(%rdi), %ymm20
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 256(%rdi), %ymm27
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm11, %ymm4
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm20, %ymm27, %ymm4
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm9 = xmm4[u,u,u,u,u,3,10],zero,zero,zero,xmm4[6,13],zero,zero,xmm4[u,u]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm4
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u],zero,zero,xmm4[1,8,15],zero,zero,xmm4[4,11,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm4, %xmm9, %xmm9
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 352(%rdi), %ymm18
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 320(%rdi), %ymm16
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm1, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm18, %ymm16, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm13 = ymm10[2,3,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm13[2],ymm10[3,4,5],ymm13[6],ymm10[7,8,9],ymm13[10],ymm10[11,12,13],ymm13[14],ymm10[15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[2,9,16,23,30,21,28,19,26,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} ymm28 = [65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm28, %ymm9, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm9 = [65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 416(%rdi), %ymm21
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 384(%rdi), %ymm22
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm13
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm13
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm13, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,u,u,u],zero,zero,zero,xmm5[6,13],zero,zero,xmm5[2,9]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[u,u,u,u,u,u,u,1,8,15],zero,zero,xmm13[4,11],zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm13, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm10, %ymm3, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm0, %zmm2
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm30 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm7, %zmm30, %zmm2
-; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 %zmm2, (%rsp) # 64-byte Spill
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm15, %ymm17, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,zero,xmm7[6,13],zero,zero,xmm7[2,9,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[1,8,15],zero,zero,xmm5[4,11],zero,zero,xmm5[u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm5, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm12, %ymm4
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm12, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm19, %ymm24, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0,1],ymm0[2],ymm7[3,4,5],ymm0[6],ymm7[7,8,9],ymm0[10],ymm7[11,12,13],ymm0[14],ymm7[15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[0,7,14,5,12,3,10,17,24,31,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm5, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm11, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm31, %ymm29, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm5[u,u,u,6,13],zero,zero,xmm5[2,9],zero,zero,zero,xmm5[u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u],zero,zero,xmm5[4,11],zero,zero,xmm5[0,7,14,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm10, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm23, %xmm2
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm2[u,u,u,u,u,u,3,10,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm6[u,u,u,u,u,u,5,12,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm13[0],xmm10[0],xmm13[1],xmm10[1],xmm13[2],xmm10[2],xmm13[3],xmm10[3]
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm10[7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = zero,zero,zero,xmm14[6,13,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm8[1,8,15],zero,zero,xmm8[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm10, %xmm13, %xmm10
-; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm10, %zmm5, %zmm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm7, %zmm26, %zmm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm27, %ymm20, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm10
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[u,u,u,u,u],zero,zero,xmm10[2,9],zero,zero,zero,xmm10[5,12,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,4,11],zero,zero,xmm7[0,7,14],zero,zero,xmm7[u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm10, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm11, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm18, %ymm16, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm13 = ymm10[2,3,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm10[0,1,2],ymm13[3],ymm10[4,5],ymm13[6],ymm10[7,8,9,10],ymm13[11],ymm10[12,13],ymm13[14],ymm10[15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[3,10,17,24,31,22,29,20,27,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm28, %ymm7, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm1, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm7[u,u,u,u,u,u,u,2,9],zero,zero,zero,xmm7[5,12],zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u],zero,zero,xmm7[0,7,14],zero,zero,xmm7[3,10]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm13, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm10, %ymm3, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm7, %zmm0, %zmm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm5, %zmm30, %zmm7
-; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm1, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm15, %ymm17, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm5[2,9],zero,zero,zero,xmm5[5,12],zero,zero,xmm5[u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,xmm5[0,7,14],zero,zero,xmm5[3,10,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm5, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm19, %ymm24, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0,1,2],ymm0[3],ymm7[4,5],ymm0[6],ymm7[7,8,9,10],ymm0[11],ymm7[12,13],ymm0[14],ymm7[15]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[1,8,15,6,13,4,11,18,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} ymm26 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm26, %ymm5, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm4, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm4, %ymm23
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm29, %ymm31, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm10
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[u,u],zero,zero,zero,xmm10[5,12],zero,zero,xmm10[1,8,15,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,0,7,14],zero,zero,xmm5[3,10],zero,zero,zero,xmm5[u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm10, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm2[u,u,u,u,u,u,4,11,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm6[u,u,u,u,u,u,6,13,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm13[0],xmm10[0],xmm13[1],xmm10[1],xmm13[2],xmm10[2],xmm13[3],xmm10[3]
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm10[7]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm8[2,9],zero,zero,zero,xmm8[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm14, %xmm4
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = zero,zero,xmm14[0,7,14,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm10, %xmm13, %xmm10
-; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm10, %zmm5, %zmm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm14 = [0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %zmm7, %zmm14, %zmm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm1, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm27, %ymm20, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm10
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[u,u,u,u,u],zero,zero,xmm10[3,10],zero,zero,zero,xmm10[6,13,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,5,12],zero,zero,xmm7[1,8,15],zero,zero,xmm7[u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm10, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm16, %ymm18, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm13 = ymm10[2,3,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm13[0],ymm10[1,2],ymm13[3],ymm10[4,5,6],ymm13[7,8],ymm10[9,10],ymm13[11],ymm10[12,13,14],ymm13[15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[4,11,18,25,16,23,30,21,28,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm28, %ymm7, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm11, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm7[u,u,u,u,u,u,u,3,10],zero,zero,zero,xmm7[6,13],zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u],zero,zero,xmm7[1,8,15],zero,zero,xmm7[4,11]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm13, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm10, %ymm3, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm7, %zmm0, %zmm25
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm5, %zmm30, %zmm25
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm11, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm15, %ymm17, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm5[3,10],zero,zero,zero,xmm5[6,13],zero,zero,xmm5[u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,xmm5[1,8,15],zero,zero,xmm5[4,11,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm5, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm1, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm19, %ymm24, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm0[0],ymm7[1,2],ymm0[3],ymm7[4,5,6],ymm0[7,8],ymm7[9,10],ymm0[11],ymm7[12,13,14],ymm0[15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[2,9,0,7,14,5,12,19,26],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm26, %ymm5, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm29, %ymm31, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm10
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[u,u],zero,zero,zero,xmm10[6,13],zero,zero,xmm10[2,9,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,1,8,15],zero,zero,xmm5[4,11],zero,zero,xmm5[u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm10, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm2[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm2[5,12]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm2, %xmm12
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm6[u,u,u,u,u,u,u,u,u,u,u,0,7,14],zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm10, %xmm13, %xmm10
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm5, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm8[3,10],zero,zero,zero,xmm8[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = zero,zero,xmm4[1,8,15,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm13, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm5, %zmm10, %zmm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %zmm7, %zmm14, %zmm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm11, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm27, %ymm20, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm7[u,u,u,u,u,6,13],zero,zero,xmm7[2,9],zero,zero,zero,xmm7[u,u]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u],zero,zero,xmm7[4,11],zero,zero,xmm7[0,7,14,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm10, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm1, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm16, %ymm18, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm13 = ymm10[2,3,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm13[0],ymm10[1,2,3],ymm13[4],ymm10[5,6],ymm13[7,8],ymm10[9,10,11],ymm13[12],ymm10[13,14],ymm13[15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[5,12,19,26,17,24,31,22,29,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm28, %ymm7, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm22, %ymm21, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm13
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[u,u,u,u,u,u,u],zero,zero,xmm13[2,9],zero,zero,zero,xmm13[5,12]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,4,11],zero,zero,xmm7[0,7,14],zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm13, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm10, %ymm3, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm7, %zmm0, %zmm26
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm5, %zmm30, %zmm26
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm17, %ymm15, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,xmm7[2,9],zero,zero,zero,xmm7[5,12,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[4,11],zero,zero,xmm5[0,7,14],zero,zero,xmm5[u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm5, %xmm2
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm23, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm20, %ymm27, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u],zero,zero,zero,xmm7[5,12],zero,zero,xmm7[1,8,15,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,0,7,14],zero,zero,xmm5[3,10],zero,zero,zero,xmm5[u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm5, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm11, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm16, %ymm18, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm7[2,3,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0],ymm10[1],ymm7[2,3],ymm10[4],ymm7[5,6,7,8],ymm10[9],ymm7[10,11],ymm10[12],ymm7[13,14,15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[6,13,20,27,18,25,16,23,30,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm28, %ymm5, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm1, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm22, %ymm21, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm10
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[u,u,u,u,u,u,u],zero,zero,xmm10[3,10],zero,zero,zero,xmm10[6,13]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,u,u,u,5,12],zero,zero,xmm5[1,8,15],zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm10, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $1, %xmm5, %ymm0, %ymm28
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm7, %ymm3, %ymm28
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm20, %ymm27, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u],zero,zero,zero,xmm7[6,13],zero,zero,xmm7[2,9,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,1,8,15],zero,zero,xmm5[4,11],zero,zero,xmm5[u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm5, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm23, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm18, %ymm16, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm7[2,3,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0],ymm10[1],ymm7[2,3,4],ymm10[5],ymm7[6,7,8],ymm10[9],ymm7[10,11,12],ymm10[13],ymm7[14,15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u],zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[0,7,14,21,28,19,26,17,24,31,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm5, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm11, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm22, %ymm21, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm5[u,u,u,u,u,u,u,6,13],zero,zero,xmm5[2,9],zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,u,u,u],zero,zero,xmm5[4,11],zero,zero,xmm5[0,7,14]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm10, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $1, %xmm5, %ymm0, %ymm30
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm7, %ymm3, %ymm30
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm1, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm29, %ymm31, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm5[u,u,2,9],zero,zero,zero,xmm5[5,12],zero,zero,xmm5[u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[0,7,14],zero,zero,xmm5[3,10,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm5, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm11, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm23
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm1, %ymm13
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm17, %ymm15, %ymm13
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm11, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm17, %ymm15, %ymm11
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm12, %xmm4
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm15 = xmm12[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm12[6,13]
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm6, %xmm3
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u,u,u,u,u,u,u,u,u,u,1,8,15],zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm14, %ymm1
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm29, %ymm1
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u],zero,zero,xmm3[3,10],zero,zero,zero,xmm3[6,13,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,5,12],zero,zero,xmm1[1,8,15],zero,zero,xmm1[u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm1, %xmm1
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 192(%rdi), %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vpbroadcastd {{.*#+}} xmm4 = [0,0,4,11,0,0,4,11,0,0,4,11,0,0,4,11]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm4, %xmm0, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm4, %xmm27
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm0, %xmm20
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 208(%rdi), %xmm10
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm10[u,u,u,u,u,u,2,9,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 240(%rdi), %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,zero,xmm0[5,12,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm0, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa 224(%rdi), %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm0[0,7,14],zero,zero,xmm0[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm0, %xmm6
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm7, %xmm5
+; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm5, %zmm1, %zmm22
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm2, %zmm4, %zmm22
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 288(%rdi), %ymm18
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 256(%rdi), %ymm16
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm18, %ymm16, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm2[u,u,u,u,u,3,10],zero,zero,zero,xmm2[6,13],zero,zero,xmm2[u,u]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u],zero,zero,xmm2[1,8,15],zero,zero,xmm2[4,11,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm2, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 352(%rdi), %ymm17
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 320(%rdi), %ymm28
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm14, %ymm7
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm17, %ymm28, %ymm7
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0,1],ymm8[2],ymm7[3,4,5],ymm8[6],ymm7[7,8,9],ymm8[10],ymm7[11,12,13],ymm8[14],ymm7[15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm8 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[2,9,16,23,30,21,28,19,26,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} ymm23 = [65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm23, %ymm2, %ymm8
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = [65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm7, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm13, %ymm12, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm15
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm15 = zero,zero,zero,xmm15[6,13],zero,zero,xmm15[2,9,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[1,8,15],zero,zero,xmm2[4,11],zero,zero,xmm2[u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm2, %xmm15, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm24, %ymm15
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm24, %ymm5
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm19, %ymm31, %ymm15
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm15 = ymm15[0,1],ymm11[2],ymm15[3,4,5],ymm11[6],ymm15[7,8,9],ymm11[10],ymm15[11,12,13],ymm11[14],ymm15[15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm15 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm15[0,7,14,5,12,3,10,17,24,31,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm15
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm29, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm2[u,u,u,6,13],zero,zero,xmm2[2,9],zero,zero,zero,xmm2[u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u],zero,zero,xmm2[4,11],zero,zero,xmm2[0,7,14,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm3[6,13,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm3, %xmm25
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm6[1,8,15],zero,zero,xmm6[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm6, %xmm26
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm2, %xmm3, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm10[u,u,u,u,u,u,3,10,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm20, %xmm1
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm1[u,u,u,u,u,u,5,12,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
+; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm24
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm15, %zmm4, %zmm24
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm14, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm13, %ymm12, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm0[2,9],zero,zero,zero,xmm0[5,12],zero,zero,xmm0[u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = zero,zero,xmm0[0,7,14],zero,zero,xmm0[3,10,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm2, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm7, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm19, %ymm31, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2],ymm11[3],ymm2[4,5],ymm11[6],ymm2[7,8,9,10],ymm11[11],ymm2[12,13],ymm11[14],ymm2[15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[1,8,15,6,13,4,11,18,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm3, %ymm0, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm3, %ymm15
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm5, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm29, %ymm21, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u],zero,zero,zero,xmm3[5,12],zero,zero,xmm3[1,8,15,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,0,7,14],zero,zero,xmm0[3,10],zero,zero,zero,xmm0[u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm27, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm3, %xmm10, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm1[u,u,u,u,u,u,6,13,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm20, %xmm5
+; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm26, %xmm1
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm26, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm1[2,9],zero,zero,zero,xmm1[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm25, %xmm4
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %xmm25, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,xmm4[0,7,14,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm6, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm3, %zmm0, %zmm25
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm20 = [0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %zmm2, %zmm20, %zmm25
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm13, %ymm12, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm0[3,10],zero,zero,zero,xmm0[6,13],zero,zero,xmm0[u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = zero,zero,xmm0[1,8,15],zero,zero,xmm0[4,11,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm2, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm14, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm19, %ymm31, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm11[0],ymm2[1,2],ymm11[3],ymm2[4,5,6],ymm11[7,8],ymm2[9,10],ymm11[11],ymm2[12,13,14],ymm11[15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[2,9,0,7,14,5,12,19,26],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm15, %ymm0, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm15, %ymm11
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm7, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm29, %ymm21, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u],zero,zero,zero,xmm2[6,13],zero,zero,xmm2[2,9,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,1,8,15],zero,zero,xmm0[4,11],zero,zero,xmm0[u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm2, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = <128,128,128,128,128,128,128,128,128,128,128,128,128,128,5,12,19,26,17,24,31,22,29,u,u,u,u,u,u,u,u,u>
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm6
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm15 = xmm5[u,u,u,u,u,u,u,u,u,u,u,0,7,14],zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm5, %xmm10
; AVX512F-ONLY-SLOW-NEXT: vpor %xmm6, %xmm15, %xmm6
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm19, %ymm24, %ymm10
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %ymm20, %ymm1, %ymm27
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm6
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm1[3,10],zero,zero,zero,xmm1[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm15 = zero,zero,xmm4[1,8,15,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm0, %xmm15, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 416(%rdi), %ymm26
+; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $2, %xmm0, %zmm6, %zmm30
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 384(%rdi), %ymm27
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %zmm3, %zmm20, %zmm30
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm7, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm26, %ymm27, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,zero,xmm3[6,13],zero,zero,xmm3[2,9]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u,1,8,15],zero,zero,xmm0[4,11],zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm8, %ymm1, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm20
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm8 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm22, %zmm8, %zmm20
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm7, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm16, %ymm18, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[5,12,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,4,11],zero,zero,xmm0[0,7,14],zero,zero,xmm0[u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm17, %ymm28, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3],ymm3[4,5],ymm6[6],ymm3[7,8,9,10],ymm6[11],ymm3[12,13],ymm6[14],ymm3[15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[3,10,17,24,31,22,29,20,27,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm23, %ymm0, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm14, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm26, %ymm27, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm0[u,u,u,u,u,u,u,2,9],zero,zero,zero,xmm0[5,12],zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u],zero,zero,xmm0[0,7,14],zero,zero,xmm0[3,10]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm6, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm3, %ymm1, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm22
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm24, %zmm8, %zmm22
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm14, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm16, %ymm18, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u],zero,zero,xmm3[3,10],zero,zero,zero,xmm3[6,13,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,5,12],zero,zero,xmm0[1,8,15],zero,zero,xmm0[u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm7, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm28, %ymm17, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm6[0],ymm3[1,2],ymm6[3],ymm3[4,5,6],ymm6[7,8],ymm3[9,10],ymm6[11],ymm3[12,13,14],ymm6[15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[4,11,18,25,16,23,30,21,28,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm23, %ymm0, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm26, %ymm27, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm0[u,u,u,u,u,u,u,3,10],zero,zero,zero,xmm0[6,13],zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u],zero,zero,xmm0[1,8,15],zero,zero,xmm0[4,11]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm6, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm3, %ymm1, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm24
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm25, %zmm8, %zmm24
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm14, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm28, %ymm17, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm0[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3],ymm3[4],ymm0[5,6],ymm3[7,8],ymm0[9,10,11],ymm3[12],ymm0[13,14],ymm3[15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %ymm2, %ymm0, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm16, %ymm18, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm2[u,u,u,u,u,6,13],zero,zero,xmm2[2,9],zero,zero,zero,xmm2[u,u]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u],zero,zero,xmm2[4,11],zero,zero,xmm2[0,7,14,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm2, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $236, %ymm23, %ymm0, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm7, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm27, %ymm26, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[5,12]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u,4,11],zero,zero,xmm0[0,7,14],zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm2, %ymm1, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm25
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm30, %zmm8, %zmm25
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm28, %ymm17, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6,7,8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14,15]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm4 = [65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm4, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm18, %ymm16, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u],zero,zero,zero,xmm3[5,12],zero,zero,xmm3[1,8,15,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,0,7,14],zero,zero,xmm2[3,10],zero,zero,zero,xmm2[u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm2, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[6,13,20,27,18,25,16,23,30,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $236, %ymm23, %ymm0, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm14, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm27, %ymm26, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,xmm3[3,10],zero,zero,zero,xmm3[6,13]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u,5,12],zero,zero,xmm0[1,8,15],zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm30
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm2, %ymm1, %ymm30
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm4, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm17, %ymm28, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4],ymm2[5],ymm0[6,7,8],ymm2[9],ymm0[10,11,12],ymm2[13],ymm0[14,15]
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm7, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm18, %ymm16, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u],zero,zero,zero,xmm3[6,13],zero,zero,xmm3[2,9,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,1,8,15],zero,zero,xmm2[4,11],zero,zero,xmm2[u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm2, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,u,u,128,128,128,128,128,128,128,128,128,0,7,14,21,28,19,26,17,24,31,u,u,u,u,u,u,u,u,u>
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %ymm5, %ymm0, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $220, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm27, %ymm26, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm0[u,u,u,u,u,u,u,6,13],zero,zero,xmm0[2,9],zero,zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u],zero,zero,xmm0[4,11],zero,zero,xmm0[0,7,14]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm23
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm2, %ymm1, %ymm23
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm7, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm12, %ymm13, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,xmm2[2,9],zero,zero,zero,xmm2[5,12,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,11],zero,zero,xmm0[0,7,14],zero,zero,xmm0[u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm2, %xmm0, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm26, %ymm27, %ymm4
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm14, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm12, %ymm13, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,xmm3[3,10],zero,zero,zero,xmm3[6,13,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[5,12],zero,zero,xmm2[1,8,15],zero,zero,xmm2[u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm2, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm3
; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm9, %ymm15
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm24, %ymm19, %ymm15
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm24, %ymm19, %ymm1
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm12, %ymm13, %ymm9
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm19, %ymm31, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm9[6,13],zero,zero,xmm9[2,9],zero,zero,zero,xmm9[u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm9, %xmm9
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm9 = zero,zero,xmm9[4,11],zero,zero,xmm9[0,7,14,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm6, %xmm9, %xmm6
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm14, %ymm12
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %ymm18, %ymm14, %ymm16
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm7, %ymm9
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm31, %ymm19, %ymm9
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm31, %ymm19, %ymm14
; AVX512F-ONLY-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm8[0],ymm10[1,2,3],ymm8[4],ymm10[5,6],ymm8[7,8],ymm10[9,10,11],ymm8[12],ymm10[13,14],ymm8[15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm14 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[3,10,1,8,15,6,13,20,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm0, %ymm2, %ymm14
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} ymm17 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm5, %ymm17, %ymm6
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm13, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,xmm5[3,10],zero,zero,zero,xmm5[6,13,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm13[5,12],zero,zero,xmm13[1,8,15],zero,zero,xmm13[u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm10, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm15[0],ymm8[1],ymm15[2,3],ymm8[4],ymm15[5,6,7,8],ymm8[9],ymm15[10,11],ymm8[12],ymm15[13,14,15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[4,11,2,9,0,7,14,21,28],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm0, %ymm5, %ymm12
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm0, %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm11[6,13],zero,zero,xmm11[2,9],zero,zero,zero,xmm11[u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm11, %xmm0
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = zero,zero,xmm0[4,11],zero,zero,xmm0[0,7,14,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm0, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm8[1],ymm1[2,3,4],ymm8[5],ymm1[6,7,8],ymm8[9],ymm1[10,11,12],ymm8[13],ymm1[14,15]
-; AVX512F-ONLY-SLOW-NEXT: vpbroadcastd {{.*#+}} xmm11 = [4,11,0,0,4,11,0,0,4,11,0,0,4,11,0,0]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[5,12,3,10,1,8,15,22,29],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm2, %ymm5, %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm11, %xmm2, %xmm1
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm15[2,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm1, %zmm6, %zmm10
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm14, %zmm1, %zmm10
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm29, %ymm31, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm7[u,u,3,10],zero,zero,zero,xmm7[6,13],zero,zero,xmm7[u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm6
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm8[0],ymm3[1,2,3],ymm8[4],ymm3[5,6],ymm8[7,8],ymm3[9,10,11],ymm8[12],ymm3[13,14],ymm8[15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[3,10,1,8,15,6,13,20,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm11, %ymm1
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm11, %ymm0, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm9[0],ymm8[1],ymm9[2,3],ymm8[4],ymm9[5,6,7,8],ymm8[9],ymm9[10,11],ymm8[12],ymm9[13,14,15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm11 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[4,11,2,9,0,7,14,21,28],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm1, %ymm2, %ymm11
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm14[0],ymm8[1],ymm14[2,3,4],ymm8[5],ymm14[6,7,8],ymm8[9],ymm14[10,11,12],ymm8[13],ymm14[14,15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm9 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[5,12,3,10,1,8,15,22,29],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, %ymm1, %ymm6, %ymm9
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm29, %ymm21, %ymm12
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm12[u,u,2,9],zero,zero,zero,xmm12[5,12],zero,zero,xmm12[u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm12, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u],zero,zero,xmm2[0,7,14],zero,zero,xmm2[3,10,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = <128,128,128,128,128,128,128,128,128,128,128,128,128,128,6,13,20,27,18,25,16,23,30,u,u,u,u,u,u,u,u,u>
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm1, %xmm12, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm10, %xmm13
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm10[u,u,u,u,u,u,u,u,u,u,u,1,8,15],zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm2, %xmm6, %xmm2
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm0, %ymm1, %ymm2
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %ymm17, %ymm7, %ymm28
+; AVX512F-ONLY-SLOW-NEXT: vpbroadcastd {{.*#+}} xmm10 = [4,11,0,0,4,11,0,0,4,11,0,0,4,11,0,0]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm29, %ymm21, %ymm15
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm29, %ymm7
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm10, %xmm8, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm14[2,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3]
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm2
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm3, %zmm0, %zmm2
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm15[u,u,3,10],zero,zero,zero,xmm15[6,13],zero,zero,xmm15[u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm15, %xmm6
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u],zero,zero,xmm6[1,8,15],zero,zero,xmm6[4,11,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm6, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $226, %ymm18, %ymm9, %ymm16
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $202, %ymm31, %ymm29, %ymm9
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm3[u,u,u,u,u,u,u,u,u,u,u,2,9],zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm4[u,u,u,u,u,u,u,u,u,u,u],zero,zero,xmm4[0,7,14]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm6, %xmm7, %xmm6
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm6, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm5, %xmm12, %xmm6
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa %xmm12, %xmm15
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm12 = xmm13[u,u,u,u,u,u,u,u,u,u,u,2,9],zero,zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm6, %xmm12, %xmm6
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm5, %ymm17, %ymm6
-; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm9, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[2,9],zero,zero,zero,xmm5[5,12,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm9[u,u,4,11],zero,zero,xmm9[0,7,14],zero,zero,xmm9[u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm7, %xmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm3[u,u,u,u,u,u,u,u,u,u,u,3,10],zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm8 = xmm4[u,u,u,u,u,u,u,u,u,u,u],zero,zero,xmm4[1,8,15]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm7, %xmm8, %xmm7
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm5, %ymm17, %ymm7
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm15[3,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm8 = xmm2[5,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm8[0],xmm5[0],xmm8[1],xmm5[1],xmm8[2],xmm5[2],xmm8[3],xmm5[3]
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm6, %zmm5
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm12, %zmm1, %zmm5
-; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm11, %xmm15, %xmm6
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm8 = xmm2[6,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm6, %zmm7, %zmm6
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm0, %zmm1, %zmm6
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm28, %zmm0, %zmm0
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm3, %ymm1, %ymm6
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[5,12,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,4,11],zero,zero,xmm7[0,7,14],zero,zero,xmm7[u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm3, %xmm7, %xmm3
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = [128,128,128,128,128,128,128,128,128,128,128,128,128,1,8,15,22,29,20,27,18,25,128,128,128,128,128,128,128,128,128,128]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm7, %xmm15, %xmm12
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[u,u,u,u,u,u,u,u,u,u,u,3,10],zero,zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm13, %xmm12, %xmm12
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %ymm3, %ymm1, %ymm12
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm14[3,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm8[5,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm13[0],xmm3[0],xmm13[1],xmm3[1],xmm13[2],xmm3[2],xmm13[3],xmm3[3]
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm3, %zmm6, %zmm3
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm11, %zmm0, %zmm3
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm10, %xmm14, %xmm6
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm8[6,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm10[0],xmm6[0],xmm10[1],xmm6[1],xmm10[2],xmm6[2],xmm10[3],xmm6[3]
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm6, %zmm12, %zmm6
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $184, %zmm9, %zmm0, %zmm6
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm30, %zmm0, %zmm0
; AVX512F-ONLY-SLOW-NEXT: movw $-512, %ax # imm = 0xFE00
; AVX512F-ONLY-SLOW-NEXT: kmovw %eax, %k1
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm0, %zmm10 {%k1}
-; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm30, %zmm0, %zmm0
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm0, %zmm5 {%k1}
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm27, %ymm1
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm1[u,u,u,u,2,9],zero,zero,zero,xmm1[5,12],zero,zero,xmm1[u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vextracti32x4 $1, %ymm27, %xmm1
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm0, %zmm2 {%k1}
+; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm23, %zmm0, %zmm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm0, %zmm3 {%k1}
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm28, %ymm1
+; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm28[2,3,0,1]
+; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7,8,9],ymm0[10],ymm1[11,12],ymm0[13],ymm1[14,15]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb %ymm7, %ymm0, %ymm0
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm16, %ymm1
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm1[u,u,u,u,2,9],zero,zero,zero,xmm1[5,12],zero,zero,xmm1[u,u,u]
+; AVX512F-ONLY-SLOW-NEXT: vextracti32x4 $1, %ymm16, %xmm1
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u],zero,zero,xmm1[0,7,14],zero,zero,xmm1[3,10,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm0, %xmm1, %xmm0
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm16, %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm16[2,3,0,1]
-; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7,8,9],ymm1[10],ymm2[11,12],ymm1[13],ymm2[14,15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[1,8,15,22,29,20,27,18,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm1
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm23, %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vextracti32x4 $1, %ymm23, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm5, %xmm1, %xmm1
+; AVX512F-ONLY-SLOW-NEXT: vpternlogq $236, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm1
+; AVX512F-ONLY-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm0
; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u],zero,zero,zero,xmm0[5,12],zero,zero,xmm0[1,8,15]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,0,7,14],zero,zero,xmm2[3,10],zero,zero,zero
-; AVX512F-ONLY-SLOW-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,0,7,14],zero,zero,xmm4[3,10],zero,zero,zero
+; AVX512F-ONLY-SLOW-NEXT: vpor %xmm0, %xmm4, %xmm0
; AVX512F-ONLY-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; AVX512F-ONLY-SLOW-NEXT: vmovdqa32 %zmm0, %zmm6 {%k1}
-; AVX512F-ONLY-SLOW-NEXT: vmovups (%rsp), %zmm0 # 64-byte Reload
-; AVX512F-ONLY-SLOW-NEXT: vmovaps %zmm0, (%rsi)
-; AVX512F-ONLY-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512F-ONLY-SLOW-NEXT: vmovaps %zmm0, (%rdx)
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm25, (%rcx)
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm26, (%r8)
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm10, (%r9)
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm20, (%rsi)
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm22, (%rdx)
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm24, (%rcx)
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm25, (%r8)
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm2, (%r9)
; AVX512F-ONLY-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm5, (%rax)
+; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm3, (%rax)
; AVX512F-ONLY-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm6, (%rax)
-; AVX512F-ONLY-SLOW-NEXT: addq $72, %rsp
; AVX512F-ONLY-SLOW-NEXT: vzeroupper
; AVX512F-ONLY-SLOW-NEXT: retq
;
; AVX512F-ONLY-FAST-LABEL: load_i8_stride7_vf64:
; AVX512F-ONLY-FAST: # %bb.0:
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} ymm26 = [65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rdi), %ymm13
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rdi), %ymm14
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rdi), %ymm10
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm26, %ymm1
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm14, %ymm13, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 (%rdi), %ymm19
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rdi), %ymm10
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 64(%rdi), %ymm30
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm0, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm0, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm10, %ymm19, %ymm1
; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm2[5,12],zero,zero,xmm2[1,8,15,u,u,u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,7,14],zero,zero,xmm1[3,10],zero,zero,zero,xmm1[u,u,u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vpor %xmm2, %xmm1, %xmm1
; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rdi), %ymm12
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 96(%rdi), %ymm27
; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm2
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm10, %ymm12, %ymm2
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 80(%rdi), %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7,8,9],ymm3[10],ymm2[11,12],ymm3[13],ymm2[14,15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm3, %ymm9
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm30, %ymm27, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 80(%rdi), %xmm9
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm9[2],ymm2[3,4],ymm9[5],ymm2[6,7,8,9],ymm9[10],ymm2[11,12],ymm9[13],ymm2[14,15]
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[6,13,4,11,2,9,16,23,30,u],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm4
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = [65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 128(%rdi), %ymm30
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 160(%rdi), %ymm31
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm11, %ymm1
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm30, %ymm31, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm12 = [65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 128(%rdi), %ymm31
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 160(%rdi), %ymm29
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm12, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm31, %ymm29, %ymm1
; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u],zero,zero,xmm2[3,10],zero,zero,zero,xmm2[6,13,u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,5,12],zero,zero,xmm1[1,8,15],zero,zero,xmm1[u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vpor %xmm2, %xmm1, %xmm1
; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [1,2,4,6,1,2,4,6]
-; AVX512F-ONLY-FAST-NEXT: # ymm3 = mem[0,1,0,1]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 192(%rdi), %ymm2
-; AVX512F-ONLY-FAST-NEXT: vpermd %ymm2, %ymm3, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,23,26,29]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7]
+; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [1,2,4,6,1,2,4,6]
+; AVX512F-ONLY-FAST-NEXT: # ymm2 = mem[0,1,0,1]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 192(%rdi), %ymm18
+; AVX512F-ONLY-FAST-NEXT: vpermd %ymm18, %ymm2, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,23,26,29]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-ONLY-FAST-NEXT: vmovdqa 240(%rdi), %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,zero,xmm3[5,12,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm3, %xmm21
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 224(%rdi), %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm3[0,7,14],zero,zero,xmm3[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm3, %xmm23
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm7, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm1, %zmm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm17 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm4, %zmm17, %zmm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 288(%rdi), %ymm18
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 256(%rdi), %ymm28
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm3[5,12,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm3, %xmm8
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 224(%rdi), %xmm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm0[0,7,14],zero,zero,xmm0[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm2, %xmm5, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm2, %zmm1, %zmm20
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm4, %zmm2, %zmm20
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 288(%rdi), %ymm11
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 256(%rdi), %ymm26
; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm4
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm18, %ymm28, %ymm4
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm4[u,u,u,u,u,3,10],zero,zero,zero,xmm4[6,13],zero,zero,xmm4[u,u]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm11, %ymm26, %ymm4
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm4[u,u,u,u,u,3,10],zero,zero,zero,xmm4[6,13],zero,zero,xmm4[u,u]
; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm4, %xmm4
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u],zero,zero,xmm4[1,8,15],zero,zero,xmm4[4,11,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm4, %xmm8, %xmm8
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 352(%rdi), %ymm16
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 320(%rdi), %ymm4
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm11, %ymm15
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm16, %ymm4, %ymm15
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm7 = ymm15[2,3,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm15[0,1],ymm7[2],ymm15[3,4,5],ymm7[6],ymm15[7,8,9],ymm7[10],ymm15[11,12,13],ymm7[14],ymm15[15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[2,9,16,23,30,21,28,19,26,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} ymm27 = [65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm27, %ymm8, %ymm7
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = [65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 416(%rdi), %ymm19
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 384(%rdi), %ymm20
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm15, %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm19, %ymm20, %ymm8
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm8, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,zero,xmm3[6,13],zero,zero,xmm3[2,9]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[u,u,u,u,u,u,u,1,8,15],zero,zero,xmm8[4,11],zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm8, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} ymm22 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm7, %ymm22, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm29 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm5, %zmm29, %zmm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm15, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm14, %ymm13, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,zero,xmm5[6,13],zero,zero,xmm5[2,9,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[1,8,15],zero,zero,xmm3[4,11],zero,zero,xmm3[u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm26, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm12, %ymm10, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0,1],ymm9[2],ymm5[3,4,5],ymm9[6],ymm5[7,8,9],ymm9[10],ymm5[11,12,13],ymm9[14],ymm5[15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm5[0,7,14,5,12,3,10,17,24,31,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm4, %xmm13
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 352(%rdi), %ymm15
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 320(%rdi), %ymm16
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm12, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm15, %ymm16, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm14 = ymm5[2,3,0,1]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0,1],ymm14[2],ymm5[3,4,5],ymm14[6],ymm5[7,8,9],ymm14[10],ymm5[11,12,13],ymm14[14],ymm5[15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm5[2,9,16,23,30,21,28,19,26,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} ymm21 = [65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm21, %ymm13, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = [65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm14, %ymm13
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm10, %ymm19, %ymm13
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm13, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,zero,xmm3[6,13],zero,zero,xmm3[2,9,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[1,8,15],zero,zero,xmm13[4,11],zero,zero,xmm13[u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm13, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm7, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm7, %ymm13
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm27, %ymm30, %ymm13
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm13 = ymm13[0,1],ymm9[2],ymm13[3,4,5],ymm9[6],ymm13[7,8,9],ymm9[10],ymm13[11,12,13],ymm9[14],ymm13[15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm13 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm13[0,7,14,5,12,3,10,17,24,31,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm13
; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm30, %ymm31, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm31, %ymm29, %ymm3
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm3[u,u,u,6,13],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u],zero,zero,xmm3[4,11],zero,zero,xmm3[0,7,14,u,u,u,u]
@@ -9734,822 +9727,849 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm7 = [1,3,4,6,1,3,4,6]
; AVX512F-ONLY-FAST-NEXT: # ymm7 = mem[0,1,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpermd %ymm2, %ymm7, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vpermd %ymm18, %ymm7, %ymm7
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,17,20,27,30]
; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm7[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm21, %xmm1
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,zero,xmm1[6,13,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm23, %xmm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,zero,xmm8[6,13,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm8, %xmm23
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm0, %xmm4
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm0[1,8,15],zero,zero,xmm0[u,u,u,u,u,u,u,u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm8, %xmm7
-; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm7, %zmm3, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm5, %zmm17, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm15, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm28, %ymm18, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm5, %xmm7
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u],zero,zero,xmm7[2,9],zero,zero,zero,xmm7[5,12,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,u,4,11],zero,zero,xmm5[0,7,14],zero,zero,xmm5[u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm5, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm7
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm16, %ymm4, %ymm7
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5],ymm8[6],ymm7[7,8,9,10],ymm8[11],ymm7[12,13],ymm8[14],ymm7[15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[3,10,17,24,31,22,29,20,27,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm27, %ymm5, %ymm7
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm11, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm19, %ymm20, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm5[u,u,u,u,u,u,u,2,9],zero,zero,zero,xmm5[5,12],zero,zero
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm5, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,u,u,u],zero,zero,xmm5[0,7,14],zero,zero,xmm5[3,10]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm8, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm7, %ymm22, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm0, %zmm17
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm3, %zmm29, %zmm17
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm11, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm14, %ymm13, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm3[2,9],zero,zero,zero,xmm3[5,12],zero,zero,xmm3[u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,xmm3[0,7,14],zero,zero,xmm3[3,10,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm15, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm12, %ymm10, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0,1,2],ymm9[3],ymm5[4,5],ymm9[6],ymm5[7,8,9,10],ymm9[11],ymm5[12,13],ymm9[14],ymm5[15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm5[1,8,15,6,13,4,11,18,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} ymm23 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm23, %ymm3, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm26, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm31, %ymm30, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm7, %zmm3, %zmm22
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm13, %zmm2, %zmm22
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm12, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm10, %ymm19, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm2[2,9],zero,zero,zero,xmm2[5,12],zero,zero,xmm2[u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm2, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,xmm2[0,7,14],zero,zero,xmm2[3,10,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm2, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm14, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm27, %ymm30, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm9[3],ymm3[4,5],ymm9[6],ymm3[7,8,9,10],ymm9[11],ymm3[12,13],ymm9[14],ymm3[15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[1,8,15,6,13,4,11,18,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} ymm17 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm17, %ymm2, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm1, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm29, %ymm31, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm2, %xmm7
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u],zero,zero,zero,xmm7[5,12],zero,zero,xmm7[1,8,15,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,0,7,14],zero,zero,xmm3[3,10],zero,zero,zero,xmm3[u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,0,7,14],zero,zero,xmm2[3,10],zero,zero,zero,xmm2[u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm2, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm7 = [1,3,5,6,1,3,5,6]
; AVX512F-ONLY-FAST-NEXT: # ymm7 = mem[0,1,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpermd %ymm2, %ymm7, %ymm2
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,21,24,31]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm0[2,9],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpermd %ymm18, %ymm7, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,21,24,31]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm7[7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm0[2,9],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,xmm1[0,7,14,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm7, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm3, %zmm2, %zmm2
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm24 = [0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm5, %zmm24, %zmm2
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm11, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm28, %ymm18, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,u],zero,zero,xmm5[3,10],zero,zero,zero,xmm5[6,13,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,5,12],zero,zero,xmm3[1,8,15],zero,zero,xmm3[u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm15, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm4, %ymm16, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm7 = ymm5[2,3,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm7[0],ymm5[1,2],ymm7[3],ymm5[4,5,6],ymm7[7,8],ymm5[9,10],ymm7[11],ymm5[12,13,14],ymm7[15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm5[4,11,18,25,16,23,30,21,28,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm27, %ymm3, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm19, %ymm20, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm3[u,u,u,u,u,u,u,3,10],zero,zero,zero,xmm3[6,13],zero,zero
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,xmm3[1,8,15],zero,zero,xmm3[4,11]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm5, %ymm22, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm21
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm2, %zmm29, %zmm21
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm23, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm23, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = zero,zero,xmm1[0,7,14,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm8, %xmm7
+; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm7, %zmm2, %zmm23
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm18 = [0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm3, %zmm18, %zmm23
; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm2
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm14, %ymm13, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm10, %ymm19, %ymm2
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm2[3,10],zero,zero,zero,xmm2[6,13],zero,zero,xmm2[u,u,u,u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm2, %xmm2
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,xmm2[1,8,15],zero,zero,xmm2[4,11,u,u,u,u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm2, %xmm2
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm11, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm12, %ymm10, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm12, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm27, %ymm30, %ymm3
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm9[0],ymm3[1,2],ymm9[3],ymm3[4,5,6],ymm9[7,8],ymm3[9,10],ymm9[11],ymm3[12,13,14],ymm9[15]
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[2,9,0,7,14,5,12,19,26],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm23, %ymm2, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm15, %ymm2
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm31, %ymm30, %ymm2
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm2, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,zero,xmm5[6,13],zero,zero,xmm5[2,9,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm17, %ymm2, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm14, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm29, %ymm31, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm2, %xmm7
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u],zero,zero,zero,xmm7[6,13],zero,zero,xmm7[2,9,u,u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,1,8,15],zero,zero,xmm2[4,11],zero,zero,xmm2[u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm2, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 208(%rdi), %xmm7
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm7[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm7[5,12]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 192(%rdi), %xmm2
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm2[u,u,u,u,u,u,u,u,u,u,u,0,7,14],zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm8, %xmm9, %xmm8
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm2, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <128,128,128,128,128,128,128,128,128,128,128,128,128,128,5,12,19,26,17,24,31,22,29,u,u,u,u,u,u,u,u,u>
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 208(%rdi), %xmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm8
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 192(%rdi), %xmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm13 = xmm0[u,u,u,u,u,u,u,u,u,u,u,0,7,14],zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm8, %xmm13, %xmm8
; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} ymm25 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm5, %ymm25, %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm0[3,10],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm9 = zero,zero,xmm1[1,8,15,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm1, %xmm0
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm9, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm8, %zmm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm3, %zmm24, %zmm5
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm7, %ymm8
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm4[3,10],zero,zero,zero,xmm4[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm13 = zero,zero,xmm1[1,8,15,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm13, %xmm7
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 416(%rdi), %ymm24
+; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $2, %xmm7, %zmm8, %zmm28
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 384(%rdi), %ymm25
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm3, %zmm18, %zmm28
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm14, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm24, %ymm25, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u],zero,zero,zero,xmm7[6,13],zero,zero,xmm7[2,9]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,1,8,15],zero,zero,xmm3[4,11],zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm5, %ymm0, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm18
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm20, %zmm5, %zmm18
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm14, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm26, %ymm11, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u],zero,zero,xmm7[2,9],zero,zero,zero,xmm7[5,12,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,4,11],zero,zero,xmm3[0,7,14],zero,zero,xmm3[u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm15, %ymm16, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5],ymm8[6],ymm7[7,8,9,10],ymm8[11],ymm7[12,13],ymm8[14],ymm7[15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[3,10,17,24,31,22,29,20,27,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm21, %ymm3, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm12, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm24, %ymm25, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm3[u,u,u,u,u,u,u,2,9],zero,zero,zero,xmm3[5,12],zero,zero
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,xmm3[0,7,14],zero,zero,xmm3[3,10]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm8, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm7, %ymm0, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm20
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm22, %zmm5, %zmm20
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm12, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm26, %ymm11, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u],zero,zero,xmm7[3,10],zero,zero,zero,xmm7[6,13,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,5,12],zero,zero,xmm3[1,8,15],zero,zero,xmm3[u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm14, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm16, %ymm15, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm8[0],ymm7[1,2],ymm8[3],ymm7[4,5,6],ymm8[7,8],ymm7[9,10],ymm8[11],ymm7[12,13,14],ymm8[15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[4,11,18,25,16,23,30,21,28,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm21, %ymm3, %ymm7
; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm28, %ymm18, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm3[u,u,u,u,u,6,13],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[u,u]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm24, %ymm25, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm3[u,u,u,u,u,u,u,3,10],zero,zero,zero,xmm3[6,13],zero,zero
; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u],zero,zero,xmm3[4,11],zero,zero,xmm3[0,7,14,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,xmm3[1,8,15],zero,zero,xmm3[4,11]
; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm8, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm11, %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm4, %ymm16, %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm9[0],ymm8[1,2,3],ymm9[4],ymm8[5,6],ymm9[7,8],ymm8[9,10,11],ymm9[12],ymm8[13,14],ymm9[15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm8 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm8[5,12,19,26,17,24,31,22,29,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm27, %ymm3, %ymm8
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm15, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm20, %ymm19, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm9
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm9[u,u,u,u,u,u,u],zero,zero,xmm9[2,9],zero,zero,zero,xmm9[5,12]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,4,11],zero,zero,xmm3[0,7,14],zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm9, %xmm3
; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm8, %ymm22, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm24
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm5, %zmm29, %zmm24
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm15, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm13, %ymm14, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,xmm5[2,9],zero,zero,zero,xmm5[5,12,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[4,11],zero,zero,xmm3[0,7,14],zero,zero,xmm3[u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm3, %xmm1
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm26, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm18, %ymm28, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm7, %ymm0, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm22
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm23, %zmm5, %zmm22
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm12, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm16, %ymm15, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm7[0],ymm3[1,2,3],ymm7[4],ymm3[5,6],ymm7[7,8],ymm3[9,10,11],ymm7[12],ymm3[13,14],ymm7[15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm26, %ymm11, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm3[u,u,u,u,u,6,13],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[u,u]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u],zero,zero,xmm3[4,11],zero,zero,xmm3[0,7,14,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $236, %ymm21, %ymm2, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm14, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm25, %ymm24, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm2, %xmm7
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u],zero,zero,xmm7[2,9],zero,zero,zero,xmm7[5,12]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,u,4,11],zero,zero,xmm2[0,7,14],zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm2, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm3, %ymm0, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm23
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm28, %zmm5, %zmm23
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm16, %ymm15, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5,6,7,8],ymm3[9],ymm2[10,11],ymm3[12],ymm2[13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm1, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm11, %ymm26, %ymm3
; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm5
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u],zero,zero,zero,xmm5[5,12],zero,zero,xmm5[1,8,15,u,u]
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,0,7,14],zero,zero,xmm3[3,10],zero,zero,zero,xmm3[u,u]
; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm4, %ymm16, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm5[2,3,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0],ymm8[1],ymm5[2,3],ymm8[4],ymm5[5,6,7,8],ymm8[9],ymm5[10,11],ymm8[12],ymm5[13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm5[6,13,20,27,18,25,16,23,30,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm27, %ymm3, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm11, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm20, %ymm19, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm8
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[u,u,u,u,u,u,u],zero,zero,xmm8[3,10],zero,zero,zero,xmm8[6,13]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,5,12],zero,zero,xmm3[1,8,15],zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm8, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $1, %xmm3, %ymm0, %ymm27
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm5, %ymm22, %ymm27
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm15, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm18, %ymm28, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[6,13,20,27,18,25,16,23,30,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $236, %ymm21, %ymm2, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm12, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm25, %ymm24, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm2, %xmm5
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,u,u,u],zero,zero,xmm5[3,10],zero,zero,zero,xmm5[6,13]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,u,5,12],zero,zero,xmm2[1,8,15],zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm2, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm4
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm3, %ymm0, %ymm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm1, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm1, %ymm21
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm15, %ymm16, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3,4],ymm3[5],ymm2[6,7,8],ymm3[9],ymm2[10,11,12],ymm3[13],ymm2[14,15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm14, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm11, %ymm26, %ymm3
; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm5
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u],zero,zero,zero,xmm5[6,13],zero,zero,xmm5[2,9,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,1,8,15],zero,zero,xmm3[4,11],zero,zero,xmm3[u,u,u]
; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm26, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm16, %ymm4, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm5[2,3,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0],ymm8[1],ymm5[2,3,4],ymm8[5],ymm5[6,7,8],ymm8[9],ymm5[10,11,12],ymm8[13],ymm5[14,15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,u],zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm5[0,7,14,21,28,19,26,17,24,31,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm20, %ymm19, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm3[u,u,u,u,u,u,u,6,13],zero,zero,xmm3[2,9],zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,xmm3[4,11],zero,zero,xmm3[0,7,14]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm8, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $1, %xmm3, %ymm0, %ymm29
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm5, %ymm22, %ymm29
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm11, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm31, %ymm30, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm3[u,u,2,9],zero,zero,zero,xmm3[5,12],zero,zero,xmm3[u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u],zero,zero,xmm3[0,7,14],zero,zero,xmm3[3,10,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm19, %ymm20, %ymm26
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm11, %ymm9
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm13, %ymm14, %ymm9
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm13, %ymm14, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm13 = xmm7[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm7[6,13]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm14 = xmm2[u,u,u,u,u,u,u,u,u,u,u,1,8,15],zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm13, %xmm14, %xmm13
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm12, %ymm10, %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %ymm18, %ymm11, %ymm28
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm15, %ymm14
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm10, %ymm12, %ymm14
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm10, %ymm12, %ymm11
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm9, %xmm10
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm10 = zero,zero,xmm10[3,10],zero,zero,zero,xmm10[6,13,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm9[5,12],zero,zero,xmm9[1,8,15],zero,zero,xmm9[u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm10, %xmm9, %xmm9
-; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm10[0],ymm8[1,2,3],ymm10[4],ymm8[5,6],ymm10[7,8],ymm8[9,10,11],ymm10[12],ymm8[13,14],ymm10[15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm8[3,10,1,8,15,6,13,20,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm23, %ymm1, %ymm12
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm3, %ymm25, %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm14[0],ymm10[1],ymm14[2,3],ymm10[4],ymm14[5,6,7,8],ymm10[9],ymm14[10,11],ymm10[12],ymm14[13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm10, %ymm13
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[4,11,2,9,0,7,14,21,28],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm23, %ymm9, %ymm10
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm6[6,13],zero,zero,xmm6[2,9],zero,zero,zero,xmm6[u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,u,u,128,128,128,128,128,128,128,128,128,0,7,14,21,28,19,26,17,24,31,u,u,u,u,u,u,u,u,u>
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm5, %ymm2, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $220, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm25, %ymm24, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm2[u,u,u,u,u,u,u,6,13],zero,zero,xmm2[2,9],zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm2, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,u],zero,zero,xmm2[4,11],zero,zero,xmm2[0,7,14]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm2, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vinserti32x4 $1, %xmm2, %ymm0, %ymm28
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm3, %ymm0, %ymm28
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm14, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm19, %ymm10, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,xmm3[2,9],zero,zero,zero,xmm3[5,12,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[4,11],zero,zero,xmm2[0,7,14],zero,zero,xmm2[u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm2, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm24, %ymm25, %ymm21
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm12, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm19, %ymm10, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,xmm7[3,10],zero,zero,zero,xmm7[6,13,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[5,12],zero,zero,xmm3[1,8,15],zero,zero,xmm3[u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm13
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm19, %ymm10, %ymm6
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm27, %ymm30, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm6[6,13],zero,zero,xmm6[2,9],zero,zero,zero,xmm6[u,u,u,u,u,u,u]
; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm6, %xmm6
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,xmm6[4,11],zero,zero,xmm6[0,7,14,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm6, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm11[0],ymm13[1],ymm11[2,3,4],ymm13[5],ymm11[6,7,8],ymm13[9],ymm11[10,11,12],ymm13[13],ymm11[14,15]
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastd {{.*#+}} xmm9 = [4,11,0,0,4,11,0,0,4,11,0,0,4,11,0,0]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[5,12,3,10,1,8,15,22,29],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm23, %ymm3, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm9, %xmm13, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm0, %xmm14
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm0[2,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm11[0],xmm3[1],xmm11[1],xmm3[2],xmm11[2],xmm3[3],xmm11[3]
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm8, %zmm8
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm6, %xmm8, %xmm8
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm12, %ymm10
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %ymm11, %ymm12, %ymm26
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm14, %ymm6
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm30, %ymm27, %ymm6
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm30, %ymm27, %ymm12
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm9, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm9[0],ymm7[1,2,3],ymm9[4],ymm7[5,6],ymm9[7,8],ymm7[9,10,11],ymm9[12],ymm7[13,14],ymm9[15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm9 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[3,10,1,8,15,6,13,20,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm17, %ymm2, %ymm9
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm6[0],ymm0[1],ymm6[2,3],ymm0[4],ymm6[5,6,7,8],ymm0[9],ymm6[10,11],ymm0[12],ymm6[13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[4,11,2,9,0,7,14,21,28],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm17, %ymm3, %ymm6
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm12[0],ymm0[1],ymm12[2,3,4],ymm0[5],ymm12[6,7,8],ymm0[9],ymm12[10,11,12],ymm0[13],ymm12[14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[5,12,3,10,1,8,15,22,29],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm17, %ymm8, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm29, %ymm31, %ymm10
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm10[u,u,2,9],zero,zero,zero,xmm10[5,12],zero,zero,xmm10[u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm10, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u],zero,zero,xmm3[0,7,14],zero,zero,xmm3[3,10,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm2, %xmm3, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <128,128,128,128,128,128,128,128,128,128,128,128,128,128,6,13,20,27,18,25,16,23,30,u,u,u,u,u,u,u,u,u>
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm0, %xmm1, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm12[u,u,u,u,u,u,u,u,u,u,u,1,8,15],zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm8, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm2, %ymm11, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %ymm15, %ymm14, %ymm16
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastd {{.*#+}} xmm8 = [4,11,0,0,4,11,0,0,4,11,0,0,4,11,0,0]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm29, %ymm31, %ymm13
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm31, %ymm29, %ymm14
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm8, %xmm0, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm15[2,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm12, %zmm3, %zmm8
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm31, %ymm30, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm5[u,u,3,10],zero,zero,zero,xmm5[6,13],zero,zero,xmm5[u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm5, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[1,8,15],zero,zero,xmm5[4,11,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm11, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %ymm16, %ymm15, %ymm4
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $202, %ymm30, %ymm31, %ymm15
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm2[u,u,u,u,u,u,u,u,u,u,u,2,9],zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm7[u,u,u,u,u,u,u,u,u,u,u],zero,zero,xmm7[0,7,14]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm11, %xmm12, %xmm11
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm9, %zmm3, %zmm2
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm13[u,u,3,10],zero,zero,zero,xmm13[6,13],zero,zero,xmm13[u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm13, %xmm10
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[u,u],zero,zero,xmm10[1,8,15],zero,zero,xmm10[4,11,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm9, %xmm10, %xmm9
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm5, %xmm1, %xmm5
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm12[u,u,u,u,u,u,u,u,u,u,u,2,9],zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm10, %xmm5
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm9, %ymm11, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm11, %ymm13
+; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm14, %xmm9
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm9[u,u],zero,zero,xmm9[2,9],zero,zero,zero,xmm9[5,12,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm14[u,u,4,11],zero,zero,xmm14[0,7,14],zero,zero,xmm14[u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm9, %xmm10, %xmm9
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = [128,128,128,128,128,128,128,128,128,128,128,128,128,1,8,15,22,29,20,27,18,25,128,128,128,128,128,128,128,128,128,128]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm10, %xmm1, %xmm11
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm12[u,u,u,u,u,u,u,u,u,u,u,3,10],zero,zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm12, %xmm11, %xmm11
+; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm5, %ymm25, %ymm11
-; AVX512F-ONLY-FAST-NEXT: vextracti128 $1, %ymm15, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[2,9],zero,zero,zero,xmm5[5,12,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm15[u,u,4,11],zero,zero,xmm15[0,7,14],zero,zero,xmm15[u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm5, %xmm12, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,u,u,u,u,u,3,10],zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,u,u,u,u],zero,zero,xmm7[1,8,15]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm2, %xmm7, %xmm2
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm5, %ymm25, %ymm2
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm14[3,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm13[5,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3]
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm11, %zmm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm10, %zmm3, %zmm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm9, %xmm14, %xmm7
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm13[6,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm9[0],xmm7[0],xmm9[1],xmm7[1],xmm9[2],xmm7[2],xmm9[3],xmm7[3]
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm2, %zmm2
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm6, %zmm3, %zmm2
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm27, %zmm0, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %ymm9, %ymm13, %ymm11
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm15[3,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm0[5,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm12[0],xmm9[0],xmm12[1],xmm9[1],xmm12[2],xmm9[2],xmm12[3],xmm9[3]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm5, %zmm5
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm6, %zmm3, %zmm5
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm8, %xmm15, %xmm6
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm0[6,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm11, %zmm6
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm7, %zmm3, %zmm6
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm3
; AVX512F-ONLY-FAST-NEXT: movw $-512, %ax # imm = 0xFE00
; AVX512F-ONLY-FAST-NEXT: kmovw %eax, %k1
-; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm3, %zmm8 {%k1}
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm29, %zmm0, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm3, %zmm2 {%k1}
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm28, %zmm0, %zmm3
; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm3, %zmm5 {%k1}
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm28, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm0[u,u,u,u,2,9],zero,zero,zero,xmm0[5,12],zero,zero,xmm0[u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vextracti32x4 $1, %ymm28, %xmm1
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u],zero,zero,xmm1[0,7,14],zero,zero,xmm1[3,10,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm1, %xmm1
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm4[2,3,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7,8,9],ymm3[10],ymm4[11,12],ymm3[13],ymm4[14,15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[1,8,15,22,29,20,27,18,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm16, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm16[2,3,0,1]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm0[0,1],ymm3[2],ymm0[3,4],ymm3[5],ymm0[6,7,8,9],ymm3[10],ymm0[11,12],ymm3[13],ymm0[14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm10, %ymm3, %ymm3
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm26, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm4 = xmm0[u,u,u,u,2,9],zero,zero,zero,xmm0[5,12],zero,zero,xmm0[u,u,u]
; AVX512F-ONLY-FAST-NEXT: vextracti32x4 $1, %ymm26, %xmm1
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u],zero,zero,zero,xmm1[5,12],zero,zero,xmm1[1,8,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u],zero,zero,xmm1[0,7,14],zero,zero,xmm1[3,10,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm4, %xmm1, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $236, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm21, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vextracti32x4 $1, %ymm21, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u],zero,zero,zero,xmm3[5,12],zero,zero,xmm3[1,8,15]
; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,0,7,14],zero,zero,xmm0[3,10],zero,zero,zero
-; AVX512F-ONLY-FAST-NEXT: vpor %xmm1, %xmm0, %xmm0
+; AVX512F-ONLY-FAST-NEXT: vpor %xmm3, %xmm0, %xmm0
; AVX512F-ONLY-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7],ymm3[8,9,10],ymm0[11,12,13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm0, %zmm2 {%k1}
-; AVX512F-ONLY-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vmovaps %zmm0, (%rsi)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm17, (%rdx)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm21, (%rcx)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm24, (%r8)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm8, (%r9)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa32 %zmm0, %zmm6 {%k1}
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm18, (%rsi)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm20, (%rdx)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm22, (%rcx)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm23, (%r8)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm2, (%r9)
; AVX512F-ONLY-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm5, (%rax)
; AVX512F-ONLY-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm2, (%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm6, (%rax)
; AVX512F-ONLY-FAST-NEXT: vzeroupper
; AVX512F-ONLY-FAST-NEXT: retq
;
; AVX512DQ-SLOW-LABEL: load_i8_stride7_vf64:
; AVX512DQ-SLOW: # %bb.0:
-; AVX512DQ-SLOW-NEXT: subq $104, %rsp
-; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} ymm28 = [65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 (%rdi), %ymm19
-; AVX512DQ-SLOW-NEXT: vmovdqa 32(%rdi), %ymm14
-; AVX512DQ-SLOW-NEXT: vmovdqa64 64(%rdi), %ymm27
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm28, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm14, %ymm19, %ymm0
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,zero,xmm1[5,12],zero,zero,xmm1[1,8,15,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,7,14],zero,zero,xmm0[3,10],zero,zero,zero,xmm0[u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm1, %xmm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = [65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 96(%rdi), %ymm26
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm1
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm27, %ymm26, %ymm1
-; AVX512DQ-SLOW-NEXT: vmovdqa 80(%rdi), %xmm2
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm2, %ymm24
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[6,13,4,11,2,9,16,23,30,u],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm2
-; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm15 = [65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 128(%rdi), %ymm31
-; AVX512DQ-SLOW-NEXT: vmovdqa64 160(%rdi), %ymm30
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm15, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm31, %ymm30, %ymm0
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u],zero,zero,xmm4[3,10],zero,zero,zero,xmm4[6,13,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,5,12],zero,zero,xmm0[1,8,15],zero,zero,xmm0[u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512DQ-SLOW-NEXT: vmovdqa 192(%rdi), %xmm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm4[u,u,u,u,u,u,4,11,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vmovdqa %xmm4, %xmm6
-; AVX512DQ-SLOW-NEXT: vmovdqa 208(%rdi), %xmm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm9 = xmm4[u,u,u,u,u,u,2,9,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm4, %xmm23
-; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3]
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm5[7]
-; AVX512DQ-SLOW-NEXT: vmovdqa 240(%rdi), %xmm12
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,zero,xmm12[5,12,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vmovdqa 224(%rdi), %xmm8
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm9 = xmm8[0,7,14],zero,zero,xmm8[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm5, %xmm9, %xmm5
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm5, %zmm0, %zmm0
-; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm11 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm2, %zmm11, %zmm0
-; AVX512DQ-SLOW-NEXT: vmovdqa64 288(%rdi), %ymm20
+; AVX512DQ-SLOW-NEXT: subq $24, %rsp
+; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm0 = [65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0]
+; AVX512DQ-SLOW-NEXT: vmovdqa (%rdi), %ymm12
+; AVX512DQ-SLOW-NEXT: vmovdqa 32(%rdi), %ymm13
+; AVX512DQ-SLOW-NEXT: vmovdqa64 64(%rdi), %ymm31
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm0, %ymm1
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm0, %ymm23
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm13, %ymm12, %ymm1
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm2[5,12],zero,zero,xmm2[1,8,15,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,7,14],zero,zero,xmm1[3,10],zero,zero,zero,xmm1[u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm2, %xmm1, %xmm1
+; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm9 = [65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 96(%rdi), %ymm28
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm2
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm31, %ymm28, %ymm2
+; AVX512DQ-SLOW-NEXT: vmovdqa 80(%rdi), %xmm11
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm11[2],ymm2[3,4],ymm11[5],ymm2[6,7,8,9],ymm11[10],ymm2[11,12],ymm11[13],ymm2[14,15]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[6,13,4,11,2,9,16,23,30,u],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-SLOW-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm2
+; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm14 = [65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535]
+; AVX512DQ-SLOW-NEXT: vmovdqa 128(%rdi), %ymm0
+; AVX512DQ-SLOW-NEXT: vmovdqa 160(%rdi), %ymm3
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm14, %ymm1
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm0, %ymm3, %ymm1
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm3, %ymm4
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm0, %ymm25
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u],zero,zero,xmm3[3,10],zero,zero,zero,xmm3[6,13,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,5,12],zero,zero,xmm1[1,8,15],zero,zero,xmm1[u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm1, %xmm1
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512DQ-SLOW-NEXT: vmovdqa 192(%rdi), %xmm0
+; AVX512DQ-SLOW-NEXT: vpbroadcastd {{.*#+}} xmm5 = [0,0,4,11,0,0,4,11,0,0,4,11,0,0,4,11]
+; AVX512DQ-SLOW-NEXT: vpshufb %xmm5, %xmm0, %xmm3
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm5, %xmm30
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm0, %xmm20
+; AVX512DQ-SLOW-NEXT: vmovdqa 208(%rdi), %xmm10
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm10[u,u,u,u,u,u,2,9,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3]
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7]
+; AVX512DQ-SLOW-NEXT: vmovdqa 240(%rdi), %xmm0
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,zero,xmm0[5,12,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vmovdqa %xmm0, %xmm3
+; AVX512DQ-SLOW-NEXT: vmovdqa 224(%rdi), %xmm0
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm0[0,7,14],zero,zero,xmm0[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vmovdqa %xmm0, %xmm5
+; AVX512DQ-SLOW-NEXT: vpor %xmm6, %xmm7, %xmm6
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm6, %zmm1, %zmm22
+; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm19 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm2, %zmm19, %zmm22
+; AVX512DQ-SLOW-NEXT: vmovdqa64 288(%rdi), %ymm18
; AVX512DQ-SLOW-NEXT: vmovdqa64 256(%rdi), %ymm17
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm5
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm20, %ymm17, %ymm5
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm9 = xmm5[u,u,u,u,u,3,10],zero,zero,zero,xmm5[6,13],zero,zero,xmm5[u,u]
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,u],zero,zero,xmm5[1,8,15],zero,zero,xmm5[4,11,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm5, %xmm9, %xmm9
-; AVX512DQ-SLOW-NEXT: vmovdqa64 352(%rdi), %ymm18
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm2
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm18, %ymm17, %ymm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm2[u,u,u,u,u,3,10],zero,zero,zero,xmm2[6,13],zero,zero,xmm2[u,u]
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u],zero,zero,xmm2[1,8,15],zero,zero,xmm2[4,11,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm6, %xmm2, %xmm2
+; AVX512DQ-SLOW-NEXT: vmovdqa64 352(%rdi), %ymm21
; AVX512DQ-SLOW-NEXT: vmovdqa64 320(%rdi), %ymm16
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm15, %ymm13
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm18, %ymm16, %ymm13
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm13[2,3,0,1]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm13[0,1],ymm10[2],ymm13[3,4,5],ymm10[6],ymm13[7,8,9],ymm10[10],ymm13[11,12,13],ymm10[14],ymm13[15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[2,9,16,23,30,21,28,19,26,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm5 = [65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535]
-; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm5, %ymm9, %ymm10
-; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm9 = [65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 416(%rdi), %ymm21
-; AVX512DQ-SLOW-NEXT: vmovdqa64 384(%rdi), %ymm22
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm13
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm13
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm13, %xmm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,u],zero,zero,zero,xmm4[6,13],zero,zero,xmm4[2,9]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[u,u,u,u,u,u,u,1,8,15],zero,zero,xmm13[4,11],zero,zero
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm13, %xmm4
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} ymm25 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0]
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm10, %ymm25, %ymm4
-; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm2
-; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm29 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm0, %zmm29, %zmm2
-; AVX512DQ-SLOW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm14, %ymm19, %ymm0
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = zero,zero,zero,xmm4[6,13],zero,zero,xmm4[2,9,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,8,15],zero,zero,xmm0[4,11],zero,zero,xmm0[u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm28, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm26, %ymm27, %ymm4
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm24, %ymm1
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1],ymm1[2],ymm4[3,4,5],ymm1[6],ymm4[7,8,9],ymm1[10],ymm4[11,12,13],ymm1[14],ymm4[15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm4[0,7,14,5,12,3,10,17,24,31,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm4
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm31, %ymm30, %ymm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm0[u,u,u,6,13],zero,zero,xmm0[2,9],zero,zero,zero,xmm0[u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u],zero,zero,xmm0[4,11],zero,zero,xmm0[0,7,14,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm0, %xmm10, %xmm0
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm14, %ymm7
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm16, %ymm7
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0,1],ymm8[2],ymm7[3,4,5],ymm8[6],ymm7[7,8,9],ymm8[10],ymm7[11,12,13],ymm8[14],ymm7[15]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm8 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[2,9,16,23,30,21,28,19,26,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} ymm24 = [65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535]
+; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm24, %ymm2, %ymm8
+; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = [65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535]
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm2
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm13, %ymm12, %ymm2
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm15
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm15 = zero,zero,zero,xmm15[6,13],zero,zero,xmm15[2,9,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[1,8,15],zero,zero,xmm2[4,11],zero,zero,xmm2[u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm2, %xmm15, %xmm2
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm23, %ymm15
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm23, %ymm29
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm28, %ymm31, %ymm15
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm15 = ymm15[0,1],ymm11[2],ymm15[3,4,5],ymm11[6],ymm15[7,8,9],ymm11[10],ymm15[11,12,13],ymm11[14],ymm15[15]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm15 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm15[0,7,14,5,12,3,10,17,24,31,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm15
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm2
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm4, %ymm6
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm25, %ymm4, %ymm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm2[u,u,u,6,13],zero,zero,xmm2[2,9],zero,zero,zero,xmm2[u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u],zero,zero,xmm2[4,11],zero,zero,xmm2[0,7,14,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX512DQ-SLOW-NEXT: vmovdqa %xmm3, %xmm4
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm3[6,13,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm5[1,8,15],zero,zero,xmm5[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm5, %xmm26
+; AVX512DQ-SLOW-NEXT: vpor %xmm2, %xmm3, %xmm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm10[u,u,u,u,u,u,3,10,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm20, %xmm1
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm1[u,u,u,u,u,u,5,12,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm23, %xmm2
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm2[u,u,u,u,u,u,3,10,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm6[u,u,u,u,u,u,5,12,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm13[0],xmm10[0],xmm13[1],xmm10[1],xmm13[2],xmm10[2],xmm13[3],xmm10[3]
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm10[7]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = zero,zero,zero,xmm12[6,13,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm8[1,8,15],zero,zero,xmm8[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm10, %xmm13, %xmm10
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm10, %zmm0, %zmm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm4, %zmm11, %zmm0
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm17, %ymm20, %ymm4
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm10
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[u,u,u,u,u],zero,zero,xmm10[2,9],zero,zero,zero,xmm10[5,12,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,4,11],zero,zero,xmm4[0,7,14],zero,zero,xmm4[u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm10, %xmm4
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm10
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm18, %ymm16, %ymm10
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm13 = ymm10[2,3,0,1]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm10[0,1,2],ymm13[3],ymm10[4,5],ymm13[6],ymm10[7,8,9,10],ymm13[11],ymm10[12,13],ymm13[14],ymm10[15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[3,10,17,24,31,22,29,20,27,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm5, %ymm4, %ymm10
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm15, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm4[u,u,u,u,u,u,u,2,9],zero,zero,zero,xmm4[5,12],zero,zero
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,u],zero,zero,xmm4[0,7,14],zero,zero,xmm4[3,10]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm13, %xmm4
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm10, %ymm25, %ymm4
-; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm3
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm0, %zmm29, %zmm3
-; AVX512DQ-SLOW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm15, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm14, %ymm19, %ymm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm0[2,9],zero,zero,zero,xmm0[5,12],zero,zero,xmm0[u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm23
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm15, %zmm19, %zmm23
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm14, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm13, %ymm12, %ymm0
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm0[2,9],zero,zero,zero,xmm0[5,12],zero,zero,xmm0[u,u,u,u,u,u,u]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = zero,zero,xmm0[0,7,14],zero,zero,xmm0[3,10,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm26, %ymm27, %ymm4
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2],ymm1[3],ymm4[4,5],ymm1[6],ymm4[7,8,9,10],ymm1[11],ymm4[12,13],ymm1[14],ymm4[15]
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm24, %ymm11
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm4[1,8,15,6,13,4,11,18,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm1, %ymm0, %ymm4
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm1, %ymm23
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm28, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm30, %ymm31, %ymm0
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm10
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[u,u],zero,zero,zero,xmm10[5,12],zero,zero,xmm10[1,8,15,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm2, %xmm0, %xmm0
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm2
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm28, %ymm31, %ymm2
+; AVX512DQ-SLOW-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2],ymm11[3],ymm2[4,5],ymm11[6],ymm2[7,8,9,10],ymm11[11],ymm2[12,13],ymm11[14],ymm2[15]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[1,8,15,6,13,4,11,18,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm3, %ymm0, %ymm2
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm3, %ymm27
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm29, %ymm0
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm25, %ymm15
+; AVX512DQ-SLOW-NEXT: vmovdqu64 %ymm25, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm6, %ymm25, %ymm0
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm6, %ymm19
+; AVX512DQ-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u],zero,zero,zero,xmm3[5,12],zero,zero,xmm3[1,8,15,u,u,u,u]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,0,7,14],zero,zero,xmm0[3,10],zero,zero,zero,xmm0[u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm0, %xmm10, %xmm0
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm2[u,u,u,u,u,u,4,11,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vmovdqa %xmm2, %xmm1
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm6[u,u,u,u,u,u,6,13,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vmovdqa %xmm6, %xmm3
-; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm13[0],xmm10[0],xmm13[1],xmm10[1],xmm13[2],xmm10[2],xmm13[3],xmm10[3]
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm10[7]
-; AVX512DQ-SLOW-NEXT: vmovdqa %xmm8, %xmm2
-; AVX512DQ-SLOW-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm8[2,9],zero,zero,zero,xmm8[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = zero,zero,xmm12[0,7,14,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm10, %xmm13, %xmm10
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm10, %zmm0, %zmm0
-; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm10 = [0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512DQ-SLOW-NEXT: vpternlogq $226, %zmm4, %zmm10, %zmm0
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm15, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm17, %ymm20, %ymm4
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm13
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[u,u,u,u,u],zero,zero,xmm13[3,10],zero,zero,zero,xmm13[6,13,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,5,12],zero,zero,xmm4[1,8,15],zero,zero,xmm4[u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm13, %xmm4
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm13
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm16, %ymm18, %ymm13
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm13[2,3,0,1]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0],ymm13[1,2],ymm6[3],ymm13[4,5,6],ymm6[7,8],ymm13[9,10],ymm6[11],ymm13[12,13,14],ymm6[15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[4,11,18,25,16,23,30,21,28,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm5, %ymm4, %ymm6
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm4[u,u,u,u,u,u,u,3,10],zero,zero,zero,xmm4[6,13],zero,zero
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,u],zero,zero,xmm4[1,8,15],zero,zero,xmm4[4,11]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm13, %xmm4
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm6, %ymm25, %ymm4
-; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm0, %zmm29, %zmm4
-; AVX512DQ-SLOW-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm14, %ymm19, %ymm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm0[3,10],zero,zero,zero,xmm0[6,13],zero,zero,xmm0[u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm30, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb %xmm3, %xmm10, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm1[u,u,u,u,u,u,6,13,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm20, %xmm6
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm20, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm26, %xmm1
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm26, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm1[2,9],zero,zero,zero,xmm1[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,xmm4[0,7,14,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm5, %xmm3
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm3, %zmm0, %zmm25
+; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm20 = [0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX512DQ-SLOW-NEXT: vpternlogq $226, %zmm2, %zmm20, %zmm25
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm13, %ymm12, %ymm0
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm0[3,10],zero,zero,zero,xmm0[6,13],zero,zero,xmm0[u,u,u,u,u,u,u]
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = zero,zero,xmm0[1,8,15],zero,zero,xmm0[4,11,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm15, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm26, %ymm27, %ymm4
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm11[0],ymm4[1,2],ymm11[3],ymm4[4,5,6],ymm11[7,8],ymm4[9,10],ymm11[11],ymm4[12,13,14],ymm11[15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm4[2,9,0,7,14,5,12,19,26],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm23, %ymm0, %ymm4
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm30, %ymm31, %ymm0
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm6
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u],zero,zero,zero,xmm6[6,13],zero,zero,xmm6[2,9,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm2, %xmm0, %xmm0
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm14, %ymm2
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm28, %ymm31, %ymm2
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm11[0],ymm2[1,2],ymm11[3],ymm2[4,5,6],ymm11[7,8],ymm2[9,10],ymm11[11],ymm2[12,13,14],ymm11[15]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[2,9,0,7,14,5,12,19,26],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm27, %ymm0, %ymm3
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm27, %ymm11
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm19, %ymm15, %ymm0
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u],zero,zero,zero,xmm2[6,13],zero,zero,xmm2[2,9,u,u,u,u,u]
; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,1,8,15],zero,zero,xmm0[4,11],zero,zero,xmm0[u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm6, %xmm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm1[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm1[5,12]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = xmm3[u,u,u,u,u,u,u,u,u,u,u,0,7,14],zero,zero
-; AVX512DQ-SLOW-NEXT: vpor %xmm6, %xmm13, %xmm6
+; AVX512DQ-SLOW-NEXT: vpor %xmm2, %xmm0, %xmm0
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX512DQ-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm6
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm2[3,10],zero,zero,zero,xmm2[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm13 = zero,zero,xmm12[1,8,15,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm0, %xmm13, %xmm0
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm0, %zmm6, %zmm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $226, %zmm4, %zmm10, %zmm0
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm17, %ymm20, %ymm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm4[u,u,u,u,u,6,13],zero,zero,xmm4[2,9],zero,zero,zero,xmm4[u,u]
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u],zero,zero,xmm4[4,11],zero,zero,xmm4[0,7,14,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm6, %xmm4, %xmm4
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm15, %ymm6
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm16, %ymm18, %ymm6
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm6[2,3,0,1]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm6 = ymm10[0],ymm6[1,2,3],ymm10[4],ymm6[5,6],ymm10[7,8],ymm6[9,10,11],ymm10[12],ymm6[13,14],ymm10[15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[5,12,19,26,17,24,31,22,29,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm5, %ymm4, %ymm6
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm22, %ymm21, %ymm4
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm10
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[u,u,u,u,u,u,u],zero,zero,xmm10[2,9],zero,zero,zero,xmm10[5,12]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,u,4,11],zero,zero,xmm4[0,7,14],zero,zero
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm10, %xmm4
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm6, %ymm25, %ymm4
-; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm24
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm0, %zmm29, %zmm24
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm28, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm20, %ymm17, %ymm0
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u],zero,zero,zero,xmm4[5,12],zero,zero,xmm4[1,8,15,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,0,7,14],zero,zero,xmm0[3,10],zero,zero,zero,xmm0[u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm16, %ymm18, %ymm4
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm4[2,3,0,1]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm6[1],ymm4[2,3],ymm6[4],ymm4[5,6,7,8],ymm6[9],ymm4[10,11],ymm6[12],ymm4[13,14,15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm4[6,13,20,27,18,25,16,23,30,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm5, %ymm0, %ymm4
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm15, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm22, %ymm21, %ymm0
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm6
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u,u,u,u,u,u],zero,zero,xmm6[3,10],zero,zero,zero,xmm6[6,13]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u,5,12],zero,zero,xmm0[1,8,15],zero,zero
-; AVX512DQ-SLOW-NEXT: vpor %xmm6, %xmm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm23
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm4, %ymm25, %ymm23
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm19, %ymm14, %ymm0
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = zero,zero,xmm4[2,9],zero,zero,zero,xmm4[5,12,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,11],zero,zero,xmm0[0,7,14],zero,zero,xmm0[u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm0, %xmm2
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm20, %ymm17, %ymm0
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u],zero,zero,zero,xmm4[6,13],zero,zero,xmm4[2,9,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,1,8,15],zero,zero,xmm0[4,11],zero,zero,xmm0[u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm28, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm18, %ymm16, %ymm4
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm4[2,3,0,1]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm6[1],ymm4[2,3,4],ymm6[5],ymm4[6,7,8],ymm6[9],ymm4[10,11,12],ymm6[13],ymm4[14,15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u],zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm4[0,7,14,21,28,19,26,17,24,31,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm4
+; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = <128,128,128,128,128,128,128,128,128,128,128,128,128,128,5,12,19,26,17,24,31,22,29,u,u,u,u,u,u,u,u,u>
+; AVX512DQ-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm5
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm15 = xmm6[u,u,u,u,u,u,u,u,u,u,u,0,7,14],zero,zero
+; AVX512DQ-SLOW-NEXT: vpor %xmm5, %xmm15, %xmm5
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512DQ-SLOW-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm5
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm1[3,10],zero,zero,zero,xmm1[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm15 = zero,zero,xmm4[1,8,15,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm0, %xmm15, %xmm0
+; AVX512DQ-SLOW-NEXT: vmovdqa64 416(%rdi), %ymm26
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $2, %xmm0, %zmm5, %zmm30
+; AVX512DQ-SLOW-NEXT: vmovdqa64 384(%rdi), %ymm27
+; AVX512DQ-SLOW-NEXT: vpternlogq $226, %zmm3, %zmm20, %zmm30
; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm22, %ymm21, %ymm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm0[u,u,u,u,u,u,u,6,13],zero,zero,xmm0[2,9],zero,zero,zero
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm26, %ymm27, %ymm0
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,zero,xmm3[6,13],zero,zero,xmm3[2,9]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u,1,8,15],zero,zero,xmm0[4,11],zero,zero
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} ymm29 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0]
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm8, %ymm29, %ymm0
+; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm20
+; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm8 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm22, %zmm8, %zmm20
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm17, %ymm18, %ymm0
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[5,12,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,4,11],zero,zero,xmm0[0,7,14],zero,zero,xmm0[u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm3
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm16, %ymm3
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm3[2,3,0,1]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3],ymm3[4,5],ymm5[6],ymm3[7,8,9,10],ymm5[11],ymm3[12,13],ymm5[14],ymm3[15]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[3,10,17,24,31,22,29,20,27,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm24, %ymm0, %ymm3
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm14, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm26, %ymm27, %ymm0
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm0[u,u,u,u,u,u,u,2,9],zero,zero,zero,xmm0[5,12],zero,zero
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u],zero,zero,xmm0[4,11],zero,zero,xmm0[0,7,14]
-; AVX512DQ-SLOW-NEXT: vpor %xmm6, %xmm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm29
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm4, %ymm25, %ymm29
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm15, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm30, %ymm31, %ymm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm0[u,u,2,9],zero,zero,zero,xmm0[5,12],zero,zero,xmm0[u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u],zero,zero,xmm0[0,7,14],zero,zero,xmm0[3,10]
+; AVX512DQ-SLOW-NEXT: vpor %xmm5, %xmm0, %xmm0
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm3, %ymm29, %ymm0
+; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm22
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm23, %zmm8, %zmm22
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm14, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm17, %ymm18, %ymm0
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u],zero,zero,xmm3[3,10],zero,zero,zero,xmm3[6,13,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,5,12],zero,zero,xmm0[1,8,15],zero,zero,xmm0[u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm3
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm16, %ymm21, %ymm3
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm3[2,3,0,1]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm5[0],ymm3[1,2],ymm5[3],ymm3[4,5,6],ymm5[7,8],ymm3[9,10],ymm5[11],ymm3[12,13,14],ymm5[15]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[4,11,18,25,16,23,30,21,28,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm24, %ymm0, %ymm3
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm26, %ymm27, %ymm0
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm0[u,u,u,u,u,u,u,3,10],zero,zero,zero,xmm0[6,13],zero,zero
; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u],zero,zero,xmm0[0,7,14],zero,zero,xmm0[3,10,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vporq %xmm4, %xmm0, %xmm25
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm10
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm22, %ymm28
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm15, %ymm13
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm19, %ymm14, %ymm13
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u],zero,zero,xmm0[1,8,15],zero,zero,xmm0[4,11]
+; AVX512DQ-SLOW-NEXT: vpor %xmm5, %xmm0, %xmm0
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm3, %ymm29, %ymm0
+; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm23
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm25, %zmm8, %zmm23
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm14, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm16, %ymm21, %ymm0
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm0[2,3,0,1]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3],ymm3[4],ymm0[5,6],ymm3[7,8],ymm0[9,10,11],ymm3[12],ymm0[13,14],ymm3[15]
+; AVX512DQ-SLOW-NEXT: vpshufb %ymm2, %ymm0, %ymm0
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm2
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm17, %ymm18, %ymm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm2[u,u,u,u,u,6,13],zero,zero,xmm2[2,9],zero,zero,zero,xmm2[u,u]
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u],zero,zero,xmm2[4,11],zero,zero,xmm2[0,7,14,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm2, %xmm2
+; AVX512DQ-SLOW-NEXT: vpternlogq $236, %ymm24, %ymm0, %ymm2
; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm19, %ymm14, %ymm7
-; AVX512DQ-SLOW-NEXT: vmovdqa %xmm1, %xmm5
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm1[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm1[6,13]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm3[u,u,u,u,u,u,u,u,u,u,u,1,8,15],zero,zero
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm6, %xmm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm26, %ymm27, %ymm10
-; AVX512DQ-SLOW-NEXT: vpternlogq $226, %ymm20, %ymm15, %ymm17
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm6
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm27, %ymm26, %ymm6
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm27, %ymm26, %ymm15
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm11[0],ymm10[1,2,3],ymm11[4],ymm10[5,6],ymm11[7,8],ymm10[9,10,11],ymm11[12],ymm10[13,14],ymm11[15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm14 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm10[3,10,1,8,15,6,13,20,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm1, %ymm2, %ymm14
-; AVX512DQ-SLOW-NEXT: vinserti32x4 $1, %xmm25, %ymm0, %ymm10
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} ymm19 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm10, %ymm19, %ymm4
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm13, %xmm10
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = zero,zero,xmm10[3,10],zero,zero,zero,xmm10[6,13,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm12 = xmm13[5,12],zero,zero,xmm13[1,8,15],zero,zero,xmm13[u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm10, %xmm12, %xmm12
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0],ymm11[1],ymm6[2,3],ymm11[4],ymm6[5,6,7,8],ymm11[9],ymm6[10,11],ymm11[12],ymm6[13,14,15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[4,11,2,9,0,7,14,21,28],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm1, %ymm12, %ymm10
-; AVX512DQ-SLOW-NEXT: vmovdqa %ymm1, %ymm2
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm7[6,13],zero,zero,xmm7[2,9],zero,zero,zero,xmm7[u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm7
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,xmm7[4,11],zero,zero,xmm7[0,7,14,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm6, %xmm7, %xmm6
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm15[0],ymm11[1],ymm15[2,3,4],ymm11[5],ymm15[6,7,8],ymm11[9],ymm15[10,11,12],ymm11[13],ymm15[14,15]
-; AVX512DQ-SLOW-NEXT: vpbroadcastd {{.*#+}} xmm12 = [4,11,0,0,4,11,0,0,4,11,0,0,4,11,0,0]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm11 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[5,12,3,10,1,8,15,22,29],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm2, %ymm6, %ymm11
-; AVX512DQ-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX512DQ-SLOW-NEXT: vpshufb %xmm12, %xmm2, %xmm1
-; AVX512DQ-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm15[2,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
-; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm1, %zmm4, %zmm7
-; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm14, %zmm1, %zmm7
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm30, %ymm31, %ymm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm0[u,u,3,10],zero,zero,zero,xmm0[6,13],zero,zero,xmm0[u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u],zero,zero,xmm0[1,8,15],zero,zero,xmm0[4,11,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm0, %xmm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $226, %ymm18, %ymm9, %ymm16
-; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm31, %ymm30, %ymm9
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm3[u,u,u,u,u,u,u,u,u,u,u,2,9],zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm5[u,u,u,u,u,u,u,u,u,u,u],zero,zero,xmm5[0,7,14]
-; AVX512DQ-SLOW-NEXT: vpor %xmm4, %xmm6, %xmm4
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm27, %ymm26, %ymm0
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[5,12]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u,4,11],zero,zero,xmm0[0,7,14],zero,zero
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm0, %ymm19, %ymm4
-; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm9, %xmm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u],zero,zero,xmm0[2,9],zero,zero,zero,xmm0[5,12,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm9[u,u,4,11],zero,zero,xmm9[0,7,14],zero,zero,xmm9[u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm0, %xmm6, %xmm0
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm3[u,u,u,u,u,u,u,u,u,u,u,3,10],zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm8 = xmm5[u,u,u,u,u,u,u,u,u,u,u],zero,zero,xmm5[1,8,15]
-; AVX512DQ-SLOW-NEXT: vpor %xmm6, %xmm8, %xmm6
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm2, %ymm29, %ymm0
+; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm25
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm30, %zmm8, %zmm25
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm16, %ymm21, %ymm0
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6,7,8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14,15]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} ymm19 = [65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0]
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm19, %ymm2
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm18, %ymm17, %ymm2
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u],zero,zero,zero,xmm3[5,12],zero,zero,xmm3[1,8,15,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,0,7,14],zero,zero,xmm2[3,10],zero,zero,zero,xmm2[u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm2, %xmm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[6,13,20,27,18,25,16,23,30,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpternlogq $236, %ymm24, %ymm0, %ymm2
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm14, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm27, %ymm26, %ymm0
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,xmm3[3,10],zero,zero,zero,xmm3[6,13]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u,5,12],zero,zero,xmm0[1,8,15],zero,zero
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm24
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm2, %ymm29, %ymm24
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm19, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm21, %ymm16, %ymm0
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4],ymm2[5],ymm0[6,7,8],ymm2[9],ymm0[10,11,12],ymm2[13],ymm0[14,15]
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm2
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm18, %ymm17, %ymm2
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u],zero,zero,zero,xmm3[6,13],zero,zero,xmm3[2,9,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,1,8,15],zero,zero,xmm2[4,11],zero,zero,xmm2[u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm2, %xmm2
+; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm8 = <u,u,u,u,128,128,128,128,128,128,128,128,128,0,7,14,21,28,19,26,17,24,31,u,u,u,u,u,u,u,u,u>
+; AVX512DQ-SLOW-NEXT: vpshufb %ymm8, %ymm0, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $220, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm2
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm27, %ymm26, %ymm0
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm0[u,u,u,u,u,u,u,6,13],zero,zero,xmm0[2,9],zero,zero,zero
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u],zero,zero,xmm0[4,11],zero,zero,xmm0[0,7,14]
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
+; AVX512DQ-SLOW-NEXT: vinserti32x4 $1, %xmm0, %ymm0, %ymm30
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm2, %ymm29, %ymm30
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm12, %ymm13, %ymm0
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,xmm2[2,9],zero,zero,zero,xmm2[5,12,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,11],zero,zero,xmm0[0,7,14],zero,zero,xmm0[u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vporq %xmm2, %xmm0, %xmm29
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm26, %ymm27, %ymm19
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm14, %ymm0
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm12, %ymm13, %ymm0
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,xmm2[3,10],zero,zero,zero,xmm2[6,13,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[5,12],zero,zero,xmm0[1,8,15],zero,zero,xmm0[u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm2, %xmm0, %xmm0
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm2
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm9, %ymm15
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm12, %ymm13, %ymm9
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm28, %ymm31, %ymm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm9[6,13],zero,zero,xmm9[2,9],zero,zero,zero,xmm9[u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm9, %xmm5
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,xmm5[4,11],zero,zero,xmm5[0,7,14,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm5, %xmm3
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm14, %ymm5
+; AVX512DQ-SLOW-NEXT: vpternlogq $226, %ymm18, %ymm14, %ymm17
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm7, %ymm9
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm31, %ymm28, %ymm9
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm31, %ymm28, %ymm14
+; AVX512DQ-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm2[1,2,3],ymm4[4],ymm2[5,6],ymm4[7,8],ymm2[9,10,11],ymm4[12],ymm2[13,14],ymm4[15]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[3,10,1,8,15,6,13,20,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-SLOW-NEXT: vmovdqa %ymm11, %ymm1
+; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm11, %ymm29, %ymm12
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm9[0],ymm4[1],ymm9[2,3],ymm4[4],ymm9[5,6,7,8],ymm4[9],ymm9[10,11],ymm4[12],ymm9[13,14,15]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm11 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[4,11,2,9,0,7,14,21,28],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm1, %ymm0, %ymm11
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm14[0],ymm4[1],ymm14[2,3,4],ymm4[5],ymm14[6,7,8],ymm4[9],ymm14[10,11,12],ymm4[13],ymm14[14,15]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm9 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[5,12,3,10,1,8,15,22,29],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-SLOW-NEXT: vpternlogq $248, %ymm1, %ymm3, %ymm9
+; AVX512DQ-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX512DQ-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm6, %ymm13, %ymm5
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm5[u,u,2,9],zero,zero,zero,xmm5[5,12],zero,zero,xmm5[u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u],zero,zero,xmm2[0,7,14],zero,zero,xmm2[3,10,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX512DQ-SLOW-NEXT: vmovdqa %xmm10, %xmm1
+; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = <128,128,128,128,128,128,128,128,128,128,128,128,128,128,6,13,20,27,18,25,16,23,30,u,u,u,u,u,u,u,u,u>
+; AVX512DQ-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm2
+; AVX512DQ-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm4[u,u,u,u,u,u,u,u,u,u,u,1,8,15],zero,zero
+; AVX512DQ-SLOW-NEXT: vpor %xmm2, %xmm3, %xmm2
; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm0, %ymm19, %ymm6
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm15[3,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm8 = xmm2[5,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
-; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm4, %zmm0
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm10, %zmm1, %zmm0
-; AVX512DQ-SLOW-NEXT: vpshufb %xmm12, %xmm15, %xmm4
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm8 = xmm2[6,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3]
-; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm6, %zmm4
-; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm11, %zmm1, %zmm4
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} ymm18 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm0, %ymm18, %ymm2
+; AVX512DQ-SLOW-NEXT: vpternlogq $226, %ymm21, %ymm7, %ymm16
+; AVX512DQ-SLOW-NEXT: vpbroadcastd {{.*#+}} xmm10 = [4,11,0,0,4,11,0,0,4,11,0,0,4,11,0,0]
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm6, %ymm13, %ymm15
+; AVX512DQ-SLOW-NEXT: vpternlogq $202, %ymm13, %ymm6, %ymm7
+; AVX512DQ-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX512DQ-SLOW-NEXT: vpshufb %xmm10, %xmm14, %xmm0
+; AVX512DQ-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm13[2,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm2
+; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm12, %zmm0, %zmm2
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm15[u,u,3,10],zero,zero,zero,xmm15[6,13],zero,zero,xmm15[u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm15, %xmm5
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[1,8,15],zero,zero,xmm5[4,11,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm5, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb %xmm8, %xmm1, %xmm5
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm8 = xmm4[u,u,u,u,u,u,u,u,u,u,u,2,9],zero,zero,zero
+; AVX512DQ-SLOW-NEXT: vpor %xmm5, %xmm8, %xmm5
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm3, %ymm18, %ymm5
+; AVX512DQ-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm3
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[5,12,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,4,11],zero,zero,xmm7[0,7,14],zero,zero,xmm7[u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm3, %xmm7, %xmm3
+; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = [128,128,128,128,128,128,128,128,128,128,128,128,128,1,8,15,22,29,20,27,18,25,128,128,128,128,128,128,128,128,128,128]
+; AVX512DQ-SLOW-NEXT: vpshufb %xmm7, %xmm1, %xmm8
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm12 = xmm4[u,u,u,u,u,u,u,u,u,u,u,3,10],zero,zero,zero
+; AVX512DQ-SLOW-NEXT: vpor %xmm12, %xmm8, %xmm8
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %ymm3, %ymm18, %ymm8
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm13[3,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm12 = xmm14[5,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm12[0],xmm3[0],xmm12[1],xmm3[1],xmm12[2],xmm3[2],xmm12[3],xmm3[3]
+; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm3, %zmm5, %zmm3
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm11, %zmm0, %zmm3
+; AVX512DQ-SLOW-NEXT: vpshufb %xmm10, %xmm13, %xmm5
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm14[6,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm10[0],xmm5[0],xmm10[1],xmm5[1],xmm10[2],xmm5[2],xmm10[3],xmm5[3]
+; AVX512DQ-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm8, %zmm5
+; AVX512DQ-SLOW-NEXT: vpternlogq $184, %zmm9, %zmm0, %zmm5
; AVX512DQ-SLOW-NEXT: movw $-512, %ax # imm = 0xFE00
; AVX512DQ-SLOW-NEXT: kmovw %eax, %k1
-; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm23, %zmm0, %zmm7 {%k1}
-; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm29, %zmm0, %zmm0 {%k1}
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm17, %ymm2
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm2[u,u,u,u,2,9],zero,zero,zero,xmm2[5,12],zero,zero,xmm2[u,u,u]
-; AVX512DQ-SLOW-NEXT: vextracti32x4 $1, %ymm17, %xmm2
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u],zero,zero,xmm2[0,7,14],zero,zero,xmm2[3,10,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm16, %ymm3
-; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm16[2,3,0,1]
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7,8,9],ymm2[10],ymm3[11,12],ymm2[13],ymm3[14,15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[1,8,15,22,29,20,27,18,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm2
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm28, %ymm3
-; AVX512DQ-SLOW-NEXT: vextracti32x4 $1, %ymm28, %xmm1
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u],zero,zero,zero,xmm1[5,12],zero,zero,xmm1[1,8,15]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,0,7,14],zero,zero,xmm3[3,10],zero,zero,zero
-; AVX512DQ-SLOW-NEXT: vpor %xmm1, %xmm3, %xmm1
-; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7],ymm2[8,9,10],ymm1[11,12,13,14,15]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm1, %zmm0, %zmm4 {%k1}
-; AVX512DQ-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
-; AVX512DQ-SLOW-NEXT: vmovaps %zmm1, (%rsi)
-; AVX512DQ-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
-; AVX512DQ-SLOW-NEXT: vmovaps %zmm1, (%rdx)
-; AVX512DQ-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
-; AVX512DQ-SLOW-NEXT: vmovaps %zmm1, (%rcx)
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm24, (%r8)
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm7, (%r9)
+; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm24, %zmm0, %zmm2 {%k1}
+; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm30, %zmm0, %zmm3 {%k1}
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm16, %ymm1
+; AVX512DQ-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm16[2,3,0,1]
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7,8,9],ymm0[10],ymm1[11,12],ymm0[13],ymm1[14,15]
+; AVX512DQ-SLOW-NEXT: vpshufb %ymm7, %ymm0, %ymm0
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm17, %ymm1
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm6 = xmm1[u,u,u,u,2,9],zero,zero,zero,xmm1[5,12],zero,zero,xmm1[u,u,u]
+; AVX512DQ-SLOW-NEXT: vextracti32x4 $1, %ymm17, %xmm1
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u],zero,zero,xmm1[0,7,14],zero,zero,xmm1[3,10,u,u,u]
+; AVX512DQ-SLOW-NEXT: vpor %xmm6, %xmm1, %xmm1
+; AVX512DQ-SLOW-NEXT: vpternlogq $236, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm1
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm19, %ymm4
+; AVX512DQ-SLOW-NEXT: vextracti32x4 $1, %ymm19, %xmm0
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u],zero,zero,zero,xmm0[5,12],zero,zero,xmm0[1,8,15]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,0,7,14],zero,zero,xmm4[3,10],zero,zero,zero
+; AVX512DQ-SLOW-NEXT: vpor %xmm0, %xmm4, %xmm0
+; AVX512DQ-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX512DQ-SLOW-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm5 {%k1}
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm20, (%rsi)
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm22, (%rdx)
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm23, (%rcx)
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm25, (%r8)
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm2, (%r9)
; AVX512DQ-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm0, (%rax)
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm3, (%rax)
; AVX512DQ-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm4, (%rax)
-; AVX512DQ-SLOW-NEXT: addq $104, %rsp
+; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm5, (%rax)
+; AVX512DQ-SLOW-NEXT: addq $24, %rsp
; AVX512DQ-SLOW-NEXT: vzeroupper
; AVX512DQ-SLOW-NEXT: retq
;
; AVX512DQ-FAST-LABEL: load_i8_stride7_vf64:
; AVX512DQ-FAST: # %bb.0:
-; AVX512DQ-FAST-NEXT: pushq %rax
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} ymm22 = [65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0]
-; AVX512DQ-FAST-NEXT: vmovdqa (%rdi), %ymm13
-; AVX512DQ-FAST-NEXT: vmovdqa 32(%rdi), %ymm14
-; AVX512DQ-FAST-NEXT: vmovdqa64 64(%rdi), %ymm17
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm22, %ymm1
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm14, %ymm13, %ymm1
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0]
+; AVX512DQ-FAST-NEXT: vmovdqa64 (%rdi), %ymm26
+; AVX512DQ-FAST-NEXT: vmovdqa 32(%rdi), %ymm11
+; AVX512DQ-FAST-NEXT: vmovdqa64 64(%rdi), %ymm30
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm0, %ymm1
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm0, %ymm7
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm11, %ymm26, %ymm1
; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm2[5,12],zero,zero,xmm2[1,8,15,u,u,u,u,u,u]
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,7,14],zero,zero,xmm1[3,10],zero,zero,zero,xmm1[u,u,u,u,u,u]
; AVX512DQ-FAST-NEXT: vpor %xmm2, %xmm1, %xmm1
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = [65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512DQ-FAST-NEXT: vmovdqa 96(%rdi), %ymm12
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm2
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm17, %ymm12, %ymm2
-; AVX512DQ-FAST-NEXT: vmovdqa 80(%rdi), %xmm10
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm10[2],ymm2[3,4],ymm10[5],ymm2[6,7,8,9],ymm10[10],ymm2[11,12],ymm10[13],ymm2[14,15]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512DQ-FAST-NEXT: vmovdqa64 96(%rdi), %ymm29
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm2
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm30, %ymm29, %ymm2
+; AVX512DQ-FAST-NEXT: vmovdqa 80(%rdi), %xmm9
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm9[2],ymm2[3,4],ymm9[5],ymm2[6,7,8,9],ymm9[10],ymm2[11,12],ymm9[13],ymm2[14,15]
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[6,13,4,11,2,9,16,23,30,u],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX512DQ-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm4
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = [65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535]
-; AVX512DQ-FAST-NEXT: vmovdqa64 128(%rdi), %ymm30
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm12 = [65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535]
+; AVX512DQ-FAST-NEXT: vmovdqa 128(%rdi), %ymm0
; AVX512DQ-FAST-NEXT: vmovdqa64 160(%rdi), %ymm31
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm11, %ymm1
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm30, %ymm31, %ymm1
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm12, %ymm1
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm0, %ymm31, %ymm1
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm0, %ymm24
; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u],zero,zero,xmm2[3,10],zero,zero,zero,xmm2[6,13,u,u,u,u]
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,5,12],zero,zero,xmm1[1,8,15],zero,zero,xmm1[u,u,u,u]
; AVX512DQ-FAST-NEXT: vpor %xmm2, %xmm1, %xmm1
; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512DQ-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [1,2,4,6,1,2,4,6]
-; AVX512DQ-FAST-NEXT: # ymm3 = mem[0,1,0,1]
-; AVX512DQ-FAST-NEXT: vmovdqa 192(%rdi), %ymm2
-; AVX512DQ-FAST-NEXT: vpermd %ymm2, %ymm3, %ymm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,23,26,29]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7]
-; AVX512DQ-FAST-NEXT: vmovdqa 240(%rdi), %xmm0
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,zero,xmm0[5,12,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vmovdqa 224(%rdi), %xmm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm3[0,7,14],zero,zero,xmm3[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm3, %xmm21
-; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm7, %xmm6
-; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm6, %zmm1, %zmm6
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm24 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm4, %zmm24, %zmm6
-; AVX512DQ-FAST-NEXT: vmovdqa64 288(%rdi), %ymm18
-; AVX512DQ-FAST-NEXT: vmovdqa64 256(%rdi), %ymm29
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm4
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm18, %ymm29, %ymm4
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm4[u,u,u,u,u,3,10],zero,zero,zero,xmm4[6,13],zero,zero,xmm4[u,u]
+; AVX512DQ-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [1,2,4,6,1,2,4,6]
+; AVX512DQ-FAST-NEXT: # ymm2 = mem[0,1,0,1]
+; AVX512DQ-FAST-NEXT: vmovdqa64 192(%rdi), %ymm18
+; AVX512DQ-FAST-NEXT: vpermd %ymm18, %ymm2, %ymm2
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,23,26,29]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX512DQ-FAST-NEXT: vmovdqa 240(%rdi), %xmm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,xmm3[5,12,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm3, %xmm8
+; AVX512DQ-FAST-NEXT: vmovdqa 224(%rdi), %xmm0
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm0[0,7,14],zero,zero,xmm0[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm2, %xmm5, %xmm2
+; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm2, %zmm1, %zmm20
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm4, %zmm2, %zmm20
+; AVX512DQ-FAST-NEXT: vmovdqa 288(%rdi), %ymm10
+; AVX512DQ-FAST-NEXT: vmovdqa64 256(%rdi), %ymm19
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm4
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm10, %ymm19, %ymm4
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm4[u,u,u,u,u,3,10],zero,zero,zero,xmm4[6,13],zero,zero,xmm4[u,u]
; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm4, %xmm4
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u],zero,zero,xmm4[1,8,15],zero,zero,xmm4[4,11,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm4, %xmm8, %xmm8
-; AVX512DQ-FAST-NEXT: vmovdqa64 352(%rdi), %ymm16
-; AVX512DQ-FAST-NEXT: vmovdqa 320(%rdi), %ymm4
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm11, %ymm15
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm16, %ymm4, %ymm15
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm7 = ymm15[2,3,0,1]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm15[0,1],ymm7[2],ymm15[3,4,5],ymm7[6],ymm15[7,8,9],ymm7[10],ymm15[11,12,13],ymm7[14],ymm15[15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[2,9,16,23,30,21,28,19,26,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} ymm27 = [65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm27, %ymm8, %ymm7
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = [65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vmovdqa64 416(%rdi), %ymm19
-; AVX512DQ-FAST-NEXT: vmovdqa64 384(%rdi), %ymm20
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm15, %ymm8
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm19, %ymm20, %ymm8
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm8, %xmm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,zero,xmm3[6,13],zero,zero,xmm3[2,9]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[u,u,u,u,u,u,u,1,8,15],zero,zero,xmm8[4,11],zero,zero
-; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm8, %xmm3
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} ymm23 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0]
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm7, %ymm23, %ymm3
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm3
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm28 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm6, %zmm28, %zmm3
-; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm15, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm14, %ymm13, %ymm3
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm6
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,zero,xmm6[6,13],zero,zero,xmm6[2,9,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[1,8,15],zero,zero,xmm3[4,11],zero,zero,xmm3[u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm3, %xmm3
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm22, %ymm6
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm12, %ymm17, %ymm6
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm10[2],ymm6[3,4,5],ymm10[6],ymm6[7,8,9],ymm10[10],ymm6[11,12,13],ymm10[14],ymm6[15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[0,7,14,5,12,3,10,17,24,31,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm6
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm30, %ymm31, %ymm3
+; AVX512DQ-FAST-NEXT: vpor %xmm5, %xmm4, %xmm13
+; AVX512DQ-FAST-NEXT: vmovdqa 352(%rdi), %ymm15
+; AVX512DQ-FAST-NEXT: vmovdqa64 320(%rdi), %ymm16
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm12, %ymm5
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm15, %ymm16, %ymm5
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm14 = ymm5[2,3,0,1]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0,1],ymm14[2],ymm5[3,4,5],ymm14[6],ymm5[7,8,9],ymm14[10],ymm5[11,12,13],ymm14[14],ymm5[15]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm5[2,9,16,23,30,21,28,19,26,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} ymm23 = [65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535]
+; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm23, %ymm13, %ymm5
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = [65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535]
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm14, %ymm13
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm11, %ymm26, %ymm13
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm13, %xmm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,zero,xmm3[6,13],zero,zero,xmm3[2,9,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[1,8,15],zero,zero,xmm13[4,11],zero,zero,xmm13[u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm13, %xmm3
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm7, %ymm1
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm7, %ymm13
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm29, %ymm30, %ymm13
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm13 = ymm13[0,1],ymm9[2],ymm13[3,4,5],ymm9[6],ymm13[7,8,9],ymm9[10],ymm13[11,12,13],ymm9[14],ymm13[15]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm13 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm13[0,7,14,5,12,3,10,17,24,31,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm13
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm24, %ymm31, %ymm3
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm3[u,u,u,6,13],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[u,u,u,u]
; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u],zero,zero,xmm3[4,11],zero,zero,xmm3[0,7,14,u,u,u,u]
@@ -10557,307 +10577,334 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512DQ-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm7 = [1,3,4,6,1,3,4,6]
; AVX512DQ-FAST-NEXT: # ymm7 = mem[0,1,0,1]
-; AVX512DQ-FAST-NEXT: vpermd %ymm2, %ymm7, %ymm7
+; AVX512DQ-FAST-NEXT: vpermd %ymm18, %ymm7, %ymm7
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,17,20,27,30]
; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm7[7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,zero,xmm0[6,13,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm21, %xmm1
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm1[1,8,15],zero,zero,xmm1[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,zero,xmm8[6,13,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm8, %xmm22
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm0, %xmm4
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm0[1,8,15],zero,zero,xmm0[u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm8, %xmm7
-; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm7, %zmm3, %zmm3
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm6, %zmm24, %zmm3
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm15, %ymm6
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm29, %ymm18, %ymm6
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm6, %xmm7
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u],zero,zero,xmm7[2,9],zero,zero,zero,xmm7[5,12,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u,u,u,u,4,11],zero,zero,xmm6[0,7,14],zero,zero,xmm6[u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm6, %xmm6
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm7
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm16, %ymm4, %ymm7
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5],ymm8[6],ymm7[7,8,9,10],ymm8[11],ymm7[12,13],ymm8[14],ymm7[15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[3,10,17,24,31,22,29,20,27,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm27, %ymm6, %ymm7
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm11, %ymm6
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm19, %ymm20, %ymm6
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm6[u,u,u,u,u,u,u,2,9],zero,zero,zero,xmm6[5,12],zero,zero
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm6, %xmm6
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u,u,u,u,u,u],zero,zero,xmm6[0,7,14],zero,zero,xmm6[3,10]
-; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm8, %xmm6
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm7, %ymm23, %ymm6
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm0, %zmm25
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm3, %zmm28, %zmm25
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm11, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm14, %ymm13, %ymm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm3[2,9],zero,zero,zero,xmm3[5,12],zero,zero,xmm3[u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,xmm3[0,7,14],zero,zero,xmm3[3,10,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm3, %xmm3
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm15, %ymm6
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm12, %ymm17, %ymm6
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0,1,2],ymm10[3],ymm6[4,5],ymm10[6],ymm6[7,8,9,10],ymm10[11],ymm6[12,13],ymm10[14],ymm6[15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[1,8,15,6,13,4,11,18,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} ymm26 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm26, %ymm3, %ymm6
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm22, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm31, %ymm30, %ymm3
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm7, %zmm3, %zmm21
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm13, %zmm2, %zmm21
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm12, %ymm2
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm11, %ymm26, %ymm2
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm2[2,9],zero,zero,zero,xmm2[5,12],zero,zero,xmm2[u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm2
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,xmm2[0,7,14],zero,zero,xmm2[3,10,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm2, %xmm2
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm14, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm29, %ymm30, %ymm3
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm9[3],ymm3[4,5],ymm9[6],ymm3[7,8,9,10],ymm9[11],ymm3[12,13],ymm9[14],ymm3[15]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[1,8,15,6,13,4,11,18,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} ymm17 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm17, %ymm2, %ymm3
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm1, %ymm2
+; AVX512DQ-FAST-NEXT: vmovdqu64 %ymm24, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm31, %ymm24, %ymm2
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm7
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u],zero,zero,zero,xmm7[5,12],zero,zero,xmm7[1,8,15,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,0,7,14],zero,zero,xmm3[3,10],zero,zero,zero,xmm3[u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,0,7,14],zero,zero,xmm2[3,10],zero,zero,zero,xmm2[u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm2, %xmm2
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512DQ-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm7 = [1,3,5,6,1,3,5,6]
; AVX512DQ-FAST-NEXT: # ymm7 = mem[0,1,0,1]
-; AVX512DQ-FAST-NEXT: vpermd %ymm2, %ymm7, %ymm2
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,21,24,31]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm1[2,9],zero,zero,zero,xmm1[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm21, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = zero,zero,xmm0[0,7,14,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpermd %ymm18, %ymm7, %ymm7
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,21,24,31]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm7[7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm0[2,9],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm7, %xmm3
-; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm3, %zmm2, %zmm2
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm24 = [0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm6, %zmm24, %zmm2
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm11, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm29, %ymm18, %ymm3
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm6
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u,u,u,u],zero,zero,xmm6[3,10],zero,zero,zero,xmm6[6,13,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,5,12],zero,zero,xmm3[1,8,15],zero,zero,xmm3[u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm3, %xmm3
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm15, %ymm6
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm4, %ymm16, %ymm6
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm7[0],ymm6[1,2],ymm7[3],ymm6[4,5,6],ymm7[7,8],ymm6[9,10],ymm7[11],ymm6[12,13,14],ymm7[15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[4,11,18,25,16,23,30,21,28,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm27, %ymm3, %ymm6
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm19, %ymm20, %ymm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm3[u,u,u,u,u,u,u,3,10],zero,zero,zero,xmm3[6,13],zero,zero
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,xmm3[1,8,15],zero,zero,xmm3[4,11]
-; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm6, %ymm23, %ymm3
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm21
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm2, %zmm28, %zmm21
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm2
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm14, %ymm13, %ymm2
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm22, %xmm1
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = zero,zero,xmm1[0,7,14,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm8, %xmm7
+; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm7, %zmm2, %zmm22
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm18 = [0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm3, %zmm18, %zmm22
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm2
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm11, %ymm26, %ymm2
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm2[3,10],zero,zero,zero,xmm2[6,13],zero,zero,xmm2[u,u,u,u,u,u,u]
; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm2
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,xmm2[1,8,15],zero,zero,xmm2[4,11,u,u,u,u,u,u,u]
; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm2, %xmm2
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm11, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm12, %ymm17, %ymm3
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm10[0],ymm3[1,2],ymm10[3],ymm3[4,5,6],ymm10[7,8],ymm3[9,10],ymm10[11],ymm3[12,13,14],ymm10[15]
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm12, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm29, %ymm30, %ymm3
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm9[0],ymm3[1,2],ymm9[3],ymm3[4,5,6],ymm9[7,8],ymm3[9,10],ymm9[11],ymm3[12,13,14],ymm9[15]
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[2,9,0,7,14,5,12,19,26],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm26, %ymm2, %ymm3
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm15, %ymm2
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm31, %ymm30, %ymm2
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm6
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u],zero,zero,zero,xmm6[6,13],zero,zero,xmm6[2,9,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm17, %ymm2, %ymm3
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm14, %ymm2
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm31, %ymm24, %ymm2
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm7
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u],zero,zero,zero,xmm7[6,13],zero,zero,xmm7[2,9,u,u,u,u,u]
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,1,8,15],zero,zero,xmm2[4,11],zero,zero,xmm2[u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm2, %xmm6
-; AVX512DQ-FAST-NEXT: vmovdqa 208(%rdi), %xmm7
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm7[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm7[5,12]
-; AVX512DQ-FAST-NEXT: vmovdqa 192(%rdi), %xmm2
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm2[u,u,u,u,u,u,u,u,u,u,u,0,7,14],zero,zero
-; AVX512DQ-FAST-NEXT: vpor %xmm8, %xmm9, %xmm8
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
+; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm2, %xmm2
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm7
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <128,128,128,128,128,128,128,128,128,128,128,128,128,128,5,12,19,26,17,24,31,22,29,u,u,u,u,u,u,u,u,u>
+; AVX512DQ-FAST-NEXT: vmovdqa 208(%rdi), %xmm0
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm8
+; AVX512DQ-FAST-NEXT: vmovdqa 192(%rdi), %xmm0
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm13 = xmm0[u,u,u,u,u,u,u,u,u,u,u,0,7,14],zero,zero
+; AVX512DQ-FAST-NEXT: vpor %xmm8, %xmm13, %xmm8
; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
-; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm6, %ymm8
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm1[3,10],zero,zero,zero,xmm1[u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm9 = zero,zero,xmm0[1,8,15,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm9, %xmm6
-; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm6, %zmm8, %zmm6
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm3, %zmm24, %zmm6
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm29, %ymm18, %ymm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm3[u,u,u,u,u,6,13],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[u,u]
+; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm7, %ymm8
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm4[3,10],zero,zero,zero,xmm4[u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm13 = zero,zero,xmm1[1,8,15,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm13, %xmm7
+; AVX512DQ-FAST-NEXT: vmovdqa64 416(%rdi), %ymm24
+; AVX512DQ-FAST-NEXT: vinserti32x4 $2, %xmm7, %zmm8, %zmm28
+; AVX512DQ-FAST-NEXT: vmovdqa64 384(%rdi), %ymm25
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm3, %zmm18, %zmm28
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm14, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm24, %ymm25, %ymm3
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u],zero,zero,zero,xmm7[6,13],zero,zero,xmm7[2,9]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,1,8,15],zero,zero,xmm3[4,11],zero,zero
+; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} ymm27 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0]
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm5, %ymm27, %ymm3
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm18
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm20, %zmm5, %zmm18
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm14, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm19, %ymm10, %ymm3
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u],zero,zero,xmm7[2,9],zero,zero,zero,xmm7[5,12,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,4,11],zero,zero,xmm3[0,7,14],zero,zero,xmm3[u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm7
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm15, %ymm16, %ymm7
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5],ymm8[6],ymm7[7,8,9,10],ymm8[11],ymm7[12,13],ymm8[14],ymm7[15]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[3,10,17,24,31,22,29,20,27,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm23, %ymm3, %ymm7
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm12, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm24, %ymm25, %ymm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm3[u,u,u,u,u,u,u,2,9],zero,zero,zero,xmm3[5,12],zero,zero
; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u],zero,zero,xmm3[4,11],zero,zero,xmm3[0,7,14,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,xmm3[0,7,14],zero,zero,xmm3[3,10]
; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm8, %xmm3
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm11, %ymm8
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm4, %ymm16, %ymm8
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm9[0],ymm8[1,2,3],ymm9[4],ymm8[5,6],ymm9[7,8],ymm8[9,10,11],ymm9[12],ymm8[13,14],ymm9[15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm8 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm8[5,12,19,26,17,24,31,22,29,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm27, %ymm3, %ymm8
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm15, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm20, %ymm19, %ymm3
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm9
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm9[u,u,u,u,u,u,u],zero,zero,xmm9[2,9],zero,zero,zero,xmm9[5,12]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,4,11],zero,zero,xmm3[0,7,14],zero,zero
-; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm9, %xmm3
; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm8, %ymm23, %ymm3
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm24
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm6, %zmm28, %zmm24
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm22, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm18, %ymm29, %ymm3
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm6
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u,u,u],zero,zero,zero,xmm6[5,12],zero,zero,xmm6[1,8,15,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,0,7,14],zero,zero,xmm3[3,10],zero,zero,zero,xmm3[u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm3, %xmm3
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm6
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm4, %ymm16, %ymm6
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm6[2,3,0,1]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0],ymm8[1],ymm6[2,3],ymm8[4],ymm6[5,6,7,8],ymm8[9],ymm6[10,11],ymm8[12],ymm6[13,14,15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[6,13,20,27,18,25,16,23,30,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm27, %ymm3, %ymm6
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm11, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm20, %ymm19, %ymm3
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm8
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[u,u,u,u,u,u,u],zero,zero,xmm8[3,10],zero,zero,zero,xmm8[6,13]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,5,12],zero,zero,xmm3[1,8,15],zero,zero
-; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm8, %xmm3
-; AVX512DQ-FAST-NEXT: vinserti32x4 $1, %xmm3, %ymm0, %ymm27
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm6, %ymm23, %ymm27
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm15, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm13, %ymm14, %ymm3
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm6
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,xmm6[2,9],zero,zero,zero,xmm6[5,12,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[4,11],zero,zero,xmm3[0,7,14],zero,zero,xmm3[u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm3, %xmm1
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm15, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm18, %ymm29, %ymm3
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm6
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u,u,u],zero,zero,zero,xmm6[6,13],zero,zero,xmm6[2,9,u,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,1,8,15],zero,zero,xmm3[4,11],zero,zero,xmm3[u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm3, %xmm3
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm22, %ymm6
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm16, %ymm4, %ymm6
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm6[2,3,0,1]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0],ymm8[1],ymm6[2,3,4],ymm8[5],ymm6[6,7,8],ymm8[9],ymm6[10,11,12],ymm8[13],ymm6[14,15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u],zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[0,7,14,21,28,19,26,17,24,31,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm6
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm20, %ymm19, %ymm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm3[u,u,u,u,u,u,u,6,13],zero,zero,xmm3[2,9],zero,zero,zero
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm7, %ymm27, %ymm3
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm20
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm21, %zmm5, %zmm20
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm12, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm19, %ymm10, %ymm3
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u],zero,zero,xmm7[3,10],zero,zero,zero,xmm7[6,13,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,5,12],zero,zero,xmm3[1,8,15],zero,zero,xmm3[u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm14, %ymm7
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm16, %ymm15, %ymm7
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm8[0],ymm7[1,2],ymm8[3],ymm7[4,5,6],ymm8[7,8],ymm7[9,10],ymm8[11],ymm7[12,13,14],ymm8[15]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[4,11,18,25,16,23,30,21,28,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm23, %ymm3, %ymm7
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm24, %ymm25, %ymm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm3[u,u,u,u,u,u,u,3,10],zero,zero,zero,xmm3[6,13],zero,zero
; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,xmm3[4,11],zero,zero,xmm3[0,7,14]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u],zero,zero,xmm3[1,8,15],zero,zero,xmm3[4,11]
; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm8, %xmm3
-; AVX512DQ-FAST-NEXT: vinserti32x4 $1, %xmm3, %ymm0, %ymm28
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm6, %ymm23, %ymm28
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm11, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm31, %ymm30, %ymm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm3[u,u,2,9],zero,zero,zero,xmm3[5,12],zero,zero,xmm3[u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm7, %ymm27, %ymm3
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm21
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm22, %zmm5, %zmm21
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm12, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm16, %ymm15, %ymm3
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm7[0],ymm3[1,2,3],ymm7[4],ymm3[5,6],ymm7[7,8],ymm3[9,10,11],ymm7[12],ymm3[13,14],ymm7[15]
+; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm2
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm19, %ymm10, %ymm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm3[u,u,u,u,u,6,13],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[u,u]
; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u],zero,zero,xmm3[0,7,14],zero,zero,xmm3[3,10,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vporq %xmm6, %xmm3, %xmm23
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm8
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm19, %ymm20, %ymm22
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm11, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm13, %ymm14, %ymm3
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm6
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm13, %ymm14, %ymm5
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm7[u,u,u,u,u,u,u,u,u,u,u],zero,zero,zero,xmm7[6,13]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm13 = xmm2[u,u,u,u,u,u,u,u,u,u,u,1,8,15],zero,zero
-; AVX512DQ-FAST-NEXT: vpor %xmm9, %xmm13, %xmm9
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm12, %ymm17, %ymm8
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %ymm18, %ymm11, %ymm29
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm15, %ymm13
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm17, %ymm12, %ymm13
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm17, %ymm12, %ymm11
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm10
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm10 = zero,zero,xmm10[3,10],zero,zero,zero,xmm10[6,13,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[5,12],zero,zero,xmm3[1,8,15],zero,zero,xmm3[u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm10, %xmm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u],zero,zero,xmm3[4,11],zero,zero,xmm3[0,7,14,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX512DQ-FAST-NEXT: vpternlogq $236, %ymm23, %ymm2, %ymm3
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm14, %ymm2
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm25, %ymm24, %ymm2
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm7
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u],zero,zero,xmm7[2,9],zero,zero,zero,xmm7[5,12]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,u,4,11],zero,zero,xmm2[0,7,14],zero,zero
+; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm2, %xmm2
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm3, %ymm27, %ymm2
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm22
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm28, %zmm5, %zmm22
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm2
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm16, %ymm15, %ymm2
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5,6,7,8],ymm3[9],ymm2[10,11],ymm3[12],ymm2[13,14,15]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0]
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm0, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm10, %ymm19, %ymm3
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm5
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u],zero,zero,zero,xmm5[5,12],zero,zero,xmm5[1,8,15,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,0,7,14],zero,zero,xmm3[3,10],zero,zero,zero,xmm3[u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm5, %xmm3, %xmm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[6,13,20,27,18,25,16,23,30,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpternlogq $236, %ymm23, %ymm2, %ymm3
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm12, %ymm2
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm25, %ymm24, %ymm2
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm5
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,u,u,u],zero,zero,xmm5[3,10],zero,zero,zero,xmm5[6,13]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,u,5,12],zero,zero,xmm2[1,8,15],zero,zero
+; AVX512DQ-FAST-NEXT: vpor %xmm5, %xmm2, %xmm2
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm1
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm3, %ymm27, %ymm1
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm0, %ymm2
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm0, %ymm23
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm15, %ymm16, %ymm2
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3,4],ymm3[5],ymm2[6,7,8],ymm3[9],ymm2[10,11,12],ymm3[13],ymm2[14,15]
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm14, %ymm3
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm10, %ymm19, %ymm3
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm3, %xmm5
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u],zero,zero,zero,xmm5[6,13],zero,zero,xmm5[2,9,u,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,1,8,15],zero,zero,xmm3[4,11],zero,zero,xmm3[u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm5, %xmm3, %xmm3
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,u,u,128,128,128,128,128,128,128,128,128,0,7,14,21,28,19,26,17,24,31,u,u,u,u,u,u,u,u,u>
+; AVX512DQ-FAST-NEXT: vpshufb %ymm5, %ymm2, %ymm2
+; AVX512DQ-FAST-NEXT: vpternlogq $220, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm3
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm2
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm25, %ymm24, %ymm2
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm2[u,u,u,u,u,u,u,6,13],zero,zero,xmm2[2,9],zero,zero,zero
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm2
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,u],zero,zero,xmm2[4,11],zero,zero,xmm2[0,7,14]
+; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm2, %xmm2
+; AVX512DQ-FAST-NEXT: vinserti32x4 $1, %xmm2, %ymm0, %ymm28
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm3, %ymm27, %ymm28
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm14, %ymm2
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm26, %ymm11, %ymm2
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,xmm3[2,9],zero,zero,zero,xmm3[5,12,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[4,11],zero,zero,xmm2[0,7,14],zero,zero,xmm2[u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vporq %xmm3, %xmm2, %xmm27
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm24, %ymm25, %ymm23
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm12, %ymm2
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm26, %ymm11, %ymm2
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,xmm3[3,10],zero,zero,zero,xmm3[6,13,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[5,12],zero,zero,xmm2[1,8,15],zero,zero,xmm2[u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm2, %xmm2
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm3
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm13
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm26, %ymm11, %ymm6
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm29, %ymm30, %ymm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm6[6,13],zero,zero,xmm6[2,9],zero,zero,zero,xmm6[u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm6, %xmm6
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,xmm6[4,11],zero,zero,xmm6[0,7,14,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm7, %xmm6, %xmm6
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm12, %ymm8
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %ymm10, %ymm12, %ymm19
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm14, %ymm7
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm30, %ymm29, %ymm7
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm30, %ymm29, %ymm12
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm9[0],ymm3[1,2,3],ymm9[4],ymm3[5,6],ymm9[7,8],ymm3[9,10,11],ymm9[12],ymm3[13,14],ymm9[15]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[3,10,1,8,15,6,13,20,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm17, %ymm27, %ymm3
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0],ymm9[1],ymm7[2,3],ymm9[4],ymm7[5,6,7,8],ymm9[9],ymm7[10,11],ymm9[12],ymm7[13,14,15]
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm9, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm9 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm7[4,11,2,9,0,7,14,21,28],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm17, %ymm2, %ymm9
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm12[0],ymm0[1],ymm12[2,3,4],ymm0[5],ymm12[6,7,8],ymm0[9],ymm12[10,11,12],ymm0[13],ymm12[14,15]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[5,12,3,10,1,8,15,22,29],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm17, %ymm6, %ymm7
; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm0[0],ymm8[1,2,3],ymm0[4],ymm8[5,6],ymm0[7,8],ymm8[9,10,11],ymm0[12],ymm8[13,14],ymm0[15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm8[3,10,1,8,15,6,13,20,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm26, %ymm1, %ymm12
-; AVX512DQ-FAST-NEXT: vinserti32x4 $1, %xmm23, %ymm0, %ymm8
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm14
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm8, %ymm1, %ymm14
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm13[0],ymm0[1],ymm13[2,3],ymm0[4],ymm13[5,6,7,8],ymm0[9],ymm13[10,11],ymm0[12],ymm13[13,14,15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm8 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm8[4,11,2,9,0,7,14,21,28],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm26, %ymm3, %ymm8
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm5[6,13],zero,zero,xmm5[2,9],zero,zero,zero,xmm5[u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm5, %xmm5
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,xmm5[4,11],zero,zero,xmm5[0,7,14,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm5, %xmm3
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm11[0],ymm0[1],ymm11[2,3,4],ymm0[5],ymm11[6,7,8],ymm0[9],ymm11[10,11,12],ymm0[13],ymm11[14,15]
-; AVX512DQ-FAST-NEXT: vpbroadcastd {{.*#+}} xmm10 = [4,11,0,0,4,11,0,0,4,11,0,0,4,11,0,0]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm9 = zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm5[5,12,3,10,1,8,15,22,29],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm26, %ymm3, %ymm9
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm31, %ymm0, %ymm8
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm8[u,u,2,9],zero,zero,zero,xmm8[5,12],zero,zero,xmm8[u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm8, %xmm6
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u],zero,zero,xmm6[0,7,14],zero,zero,xmm6[3,10,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm2, %xmm6, %xmm2
+; AVX512DQ-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <128,128,128,128,128,128,128,128,128,128,128,128,128,128,6,13,20,27,18,25,16,23,30,u,u,u,u,u,u,u,u,u>
+; AVX512DQ-FAST-NEXT: vpshufb %xmm6, %xmm4, %xmm6
+; AVX512DQ-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm12[u,u,u,u,u,u,u,u,u,u,u,1,8,15],zero,zero
+; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm8, %xmm6
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm8
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm2, %ymm11, %ymm8
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %ymm15, %ymm14, %ymm16
+; AVX512DQ-FAST-NEXT: vpbroadcastd {{.*#+}} xmm6 = [4,11,0,0,4,11,0,0,4,11,0,0,4,11,0,0]
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm31, %ymm0, %ymm13
+; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm0, %ymm31, %ymm14
; AVX512DQ-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX512DQ-FAST-NEXT: vpshufb %xmm10, %xmm0, %xmm3
-; AVX512DQ-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm13[2,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm14, %zmm5
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm12, %zmm3, %zmm5
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm31, %ymm30, %ymm6
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm6[u,u,3,10],zero,zero,zero,xmm6[6,13],zero,zero,xmm6[u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm6, %xmm6
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u],zero,zero,xmm6[1,8,15],zero,zero,xmm6[4,11,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm11, %xmm6
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %ymm16, %ymm15, %ymm4
-; AVX512DQ-FAST-NEXT: vpternlogq $202, %ymm30, %ymm31, %ymm15
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm11 = xmm2[u,u,u,u,u,u,u,u,u,u,u,2,9],zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm7[u,u,u,u,u,u,u,u,u,u,u],zero,zero,xmm7[0,7,14]
-; AVX512DQ-FAST-NEXT: vpor %xmm11, %xmm12, %xmm11
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
+; AVX512DQ-FAST-NEXT: vpshufb %xmm6, %xmm0, %xmm2
+; AVX512DQ-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm15[2,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3]
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm8, %zmm2
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm8 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm3, %zmm8, %zmm2
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm13[u,u,3,10],zero,zero,zero,xmm13[6,13],zero,zero,xmm13[u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm13, %xmm10
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm10[u,u],zero,zero,xmm10[1,8,15],zero,zero,xmm10[4,11,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm10, %xmm3
+; AVX512DQ-FAST-NEXT: vpshufb %xmm5, %xmm4, %xmm5
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm12[u,u,u,u,u,u,u,u,u,u,u,2,9],zero,zero,zero
+; AVX512DQ-FAST-NEXT: vpor %xmm5, %xmm10, %xmm5
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm3, %ymm11, %ymm5
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm11, %ymm13
+; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm14, %xmm3
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u],zero,zero,xmm3[2,9],zero,zero,zero,xmm3[5,12,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm14[u,u,4,11],zero,zero,xmm14[0,7,14],zero,zero,xmm14[u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm10, %xmm3
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = [128,128,128,128,128,128,128,128,128,128,128,128,128,1,8,15,22,29,20,27,18,25,128,128,128,128,128,128,128,128,128,128]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm10, %xmm4, %xmm11
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm12[u,u,u,u,u,u,u,u,u,u,u,3,10],zero,zero,zero
+; AVX512DQ-FAST-NEXT: vpor %xmm12, %xmm11, %xmm11
+; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm6, %ymm1, %ymm11
-; AVX512DQ-FAST-NEXT: vextracti128 $1, %ymm15, %xmm6
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u],zero,zero,xmm6[2,9],zero,zero,zero,xmm6[5,12,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm15[u,u,4,11],zero,zero,xmm15[0,7,14],zero,zero,xmm15[u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm12, %xmm6
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,u,u,u,u,u,3,10],zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,u,u,u,u],zero,zero,xmm7[1,8,15]
-; AVX512DQ-FAST-NEXT: vpor %xmm2, %xmm7, %xmm2
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm6, %ymm1, %ymm2
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm13[3,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm0[5,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm11, %zmm6
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm8, %zmm3, %zmm6
-; AVX512DQ-FAST-NEXT: vpshufb %xmm10, %xmm13, %xmm7
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm0[6,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3]
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm2, %zmm2
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm9, %zmm3, %zmm2
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %ymm3, %ymm13, %ymm11
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm15[3,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm0[5,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm12[0],xmm3[0],xmm12[1],xmm3[1],xmm12[2],xmm3[2],xmm12[3],xmm3[3]
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm5, %zmm3
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm9, %zmm8, %zmm3
+; AVX512DQ-FAST-NEXT: vpshufb %xmm6, %xmm15, %xmm5
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm0[6,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm11, %zmm5
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm7, %zmm8, %zmm5
; AVX512DQ-FAST-NEXT: movw $-512, %ax # imm = 0xFE00
; AVX512DQ-FAST-NEXT: kmovw %eax, %k1
-; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm27, %zmm0, %zmm5 {%k1}
-; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm28, %zmm0, %zmm6 {%k1}
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm29, %ymm0
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm0[u,u,u,u,2,9],zero,zero,zero,xmm0[5,12],zero,zero,xmm0[u,u,u]
-; AVX512DQ-FAST-NEXT: vextracti32x4 $1, %ymm29, %xmm1
+; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm1, %zmm0, %zmm2 {%k1}
+; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm28, %zmm0, %zmm3 {%k1}
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm16, %ymm0
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm6 = ymm16[2,3,0,1]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm0[0,1],ymm6[2],ymm0[3,4],ymm6[5],ymm0[6,7,8,9],ymm6[10],ymm0[11,12],ymm6[13],ymm0[14,15]
+; AVX512DQ-FAST-NEXT: vpshufb %ymm10, %ymm4, %ymm4
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm19, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm0[u,u,u,u,2,9],zero,zero,zero,xmm0[5,12],zero,zero,xmm0[u,u,u]
+; AVX512DQ-FAST-NEXT: vextracti32x4 $1, %ymm19, %xmm1
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u],zero,zero,xmm1[0,7,14],zero,zero,xmm1[3,10,u,u,u]
-; AVX512DQ-FAST-NEXT: vpor %xmm3, %xmm1, %xmm1
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm4[2,3,0,1]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7,8,9],ymm3[10],ymm4[11,12],ymm3[13],ymm4[14,15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[1,8,15,22,29,20,27,18,25],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm3
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm22, %ymm0
-; AVX512DQ-FAST-NEXT: vextracti32x4 $1, %ymm22, %xmm1
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u],zero,zero,zero,xmm1[5,12],zero,zero,xmm1[1,8,15]
+; AVX512DQ-FAST-NEXT: vpor %xmm6, %xmm1, %xmm1
+; AVX512DQ-FAST-NEXT: vpternlogq $236, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm1
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm23, %ymm0
+; AVX512DQ-FAST-NEXT: vextracti32x4 $1, %ymm23, %xmm4
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u],zero,zero,zero,xmm4[5,12],zero,zero,xmm4[1,8,15]
; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,0,7,14],zero,zero,xmm0[3,10],zero,zero,zero
-; AVX512DQ-FAST-NEXT: vpor %xmm1, %xmm0, %xmm0
+; AVX512DQ-FAST-NEXT: vpor %xmm4, %xmm0, %xmm0
; AVX512DQ-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7],ymm3[8,9,10],ymm0[11,12,13,14,15]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
-; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm2 {%k1}
-; AVX512DQ-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vmovaps %zmm0, (%rsi)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm25, (%rdx)
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX512DQ-FAST-NEXT: vinserti32x8 $1, %ymm0, %zmm0, %zmm5 {%k1}
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm18, (%rsi)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm20, (%rdx)
; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm21, (%rcx)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm24, (%r8)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm5, (%r9)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm22, (%r8)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm2, (%r9)
; AVX512DQ-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm6, (%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm3, (%rax)
; AVX512DQ-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm2, (%rax)
-; AVX512DQ-FAST-NEXT: popq %rax
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm5, (%rax)
; AVX512DQ-FAST-NEXT: vzeroupper
; AVX512DQ-FAST-NEXT: retq
;
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 0bb8e17e0ee484..b01532ff900b50 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
@@ -5318,289 +5318,288 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX512F-FAST-LABEL: store_i16_stride7_vf32:
; AVX512F-FAST: # %bb.0:
-; AVX512F-FAST-NEXT: subq $248, %rsp
+; AVX512F-FAST-NEXT: subq $232, %rsp
; AVX512F-FAST-NEXT: vmovdqa (%rcx), %ymm1
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128]
; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm1, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa %ymm1, %ymm10
+; AVX512F-FAST-NEXT: vmovdqa %ymm1, %ymm9
; AVX512F-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT: vmovdqa (%rdx), %ymm4
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u,u,u,u,u>
; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm3
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm18
+; AVX512F-FAST-NEXT: vmovdqa %ymm4, %ymm11
; AVX512F-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT: vpor %ymm2, %ymm3, %ymm2
; AVX512F-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa (%rsi), %ymm4
-; AVX512F-FAST-NEXT: vmovdqa 32(%rsi), %ymm11
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128,128,128,128,128]
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm4, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa %ymm4, %ymm9
-; AVX512F-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-FAST-NEXT: vmovdqa (%rsi), %ymm8
+; AVX512F-FAST-NEXT: vmovdqa 32(%rsi), %ymm10
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = [128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128,128,128,128,128]
+; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm8, %ymm2
+; AVX512F-FAST-NEXT: vmovdqu %ymm8, (%rsp) # 32-byte Spill
; AVX512F-FAST-NEXT: vmovdqa (%rdi), %ymm6
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = <12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u,u,u,u,u,16,17,18,19>
-; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm6, %ymm4
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm6, %ymm16
+; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm6, %ymm3
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm6, %ymm18
; AVX512F-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vpor %ymm2, %ymm4, %ymm2
+; AVX512F-FAST-NEXT: vpor %ymm2, %ymm3, %ymm2
; AVX512F-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT: vmovdqa (%r9), %ymm13
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [128,128,128,128,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128]
; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm13, %ymm6
; AVX512F-FAST-NEXT: vmovdqa (%r8), %ymm14
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <u,u,u,u,u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm14, %ymm7
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm14, %ymm7
; AVX512F-FAST-NEXT: vporq %ymm6, %ymm7, %ymm25
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm11, %ymm6
-; AVX512F-FAST-NEXT: vmovdqa 32(%rdi), %ymm8
-; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm8, %ymm5
-; AVX512F-FAST-NEXT: vpor %ymm6, %ymm5, %ymm3
-; AVX512F-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa 32(%rcx), %ymm3
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm3, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm29
+; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm10, %ymm4
+; AVX512F-FAST-NEXT: vmovdqa 32(%rdi), %ymm6
+; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm6, %ymm5
+; AVX512F-FAST-NEXT: vpor %ymm4, %ymm5, %ymm4
+; AVX512F-FAST-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-FAST-NEXT: vmovdqa 32(%rcx), %ymm4
+; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm31
; AVX512F-FAST-NEXT: vmovdqa 32(%rdx), %ymm12
; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm12, %ymm1
; AVX512F-FAST-NEXT: vpor %ymm0, %ymm1, %ymm0
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vmovdqa 32(%r9), %ymm15
; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm15, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa 32(%r8), %ymm6
-; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm6, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa 32(%r8), %ymm4
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm4, %ymm1
; AVX512F-FAST-NEXT: vporq %ymm0, %ymm1, %ymm21
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29>
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm11, %ymm1
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm8[3,3,3,3,7,7,7,7]
+; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm10, %ymm1
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm6[3,3,3,3,7,7,7,7]
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5],ymm1[6],ymm2[7,8,9,10],ymm1[11],ymm2[12,13],ymm1[14],ymm2[15]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm11[10,11,8,9,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,u,u,u,u,26,27,24,25,u,u,u,u]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm11, %ymm30
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm8[2,2,2,2,6,6,6,6]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm2[1],ymm4[2,3],ymm2[4],ymm4[5,6,7,8],ymm2[9],ymm4[10,11],ymm2[12],ymm4[13,14,15]
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm24 = <u,2,2,3,10,u,11,u>
-; AVX512F-FAST-NEXT: vpermi2q %zmm1, %zmm2, %zmm24
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm10[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm18[3,3,3,3,7,7,7,7]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm10[10,11,8,9,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,u,u,u,u,26,27,24,25,u,u,u,u]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm10, %ymm29
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm3 = ymm6[2,2,2,2,6,6,6,6]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5,6,7,8],ymm2[9],ymm3[10,11],ymm2[12],ymm3[13,14,15]
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm22 = <u,2,2,3,10,u,11,u>
+; AVX512F-FAST-NEXT: vpermi2q %zmm1, %zmm2, %zmm22
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm9[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm11[3,3,3,3,7,7,7,7]
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6,7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13,14,15]
; AVX512F-FAST-NEXT: vmovdqa 32(%rcx), %xmm10
-; AVX512F-FAST-NEXT: vmovdqa 32(%rdx), %xmm5
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm5[0],xmm10[0],xmm5[1],xmm10[1],xmm5[2],xmm10[2],xmm5[3],xmm10[3]
+; AVX512F-FAST-NEXT: vmovdqa 32(%rdx), %xmm11
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[0,1,2,3,0,1,2,3,6,7,4,5,6,7,4,5]
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm23 = <u,2,2,3,8,u,9,u>
; AVX512F-FAST-NEXT: vpermi2q %zmm2, %zmm1, %zmm23
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm9, %ymm0
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm16[3,3,3,3,7,7,7,7]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5],ymm0[6],ymm1[7,8,9,10],ymm0[11],ymm1[12,13],ymm0[14],ymm1[15]
-; AVX512F-FAST-NEXT: vmovdqa 32(%rdi), %xmm1
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm11 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
-; AVX512F-FAST-NEXT: vmovdqa 32(%rsi), %xmm2
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm4, %xmm4
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm20 = [2,1,3,3,8,8,9,9]
-; AVX512F-FAST-NEXT: vpermi2q %zmm4, %zmm0, %zmm20
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm14[3,3,3,3,7,7,7,7]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm0[0,1],ymm4[2],ymm0[3,4],ymm4[5],ymm0[6,7,8,9],ymm4[10],ymm0[11,12],ymm4[13],ymm0[14,15]
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm22 = [2,2,2,3,8,8,8,9]
-; AVX512F-FAST-NEXT: vmovdqa 32(%r9), %xmm0
-; AVX512F-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa 32(%r8), %xmm3
-; AVX512F-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm8, %ymm0
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm18[3,3,3,3,7,7,7,7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0,1,2],ymm0[3],ymm1[4,5],ymm0[6],ymm1[7,8,9,10],ymm0[11],ymm1[12,13],ymm0[14],ymm1[15]
+; AVX512F-FAST-NEXT: vmovdqa 32(%rdi), %xmm0
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
+; AVX512F-FAST-NEXT: vmovdqa 32(%rsi), %xmm1
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm3, %xmm3
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm5, %xmm30
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm19 = [2,1,3,3,8,8,9,9]
+; AVX512F-FAST-NEXT: vpermi2q %zmm3, %zmm2, %zmm19
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm3 = ymm14[3,3,3,3,7,7,7,7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7,8,9],ymm3[10],ymm2[11,12],ymm3[13],ymm2[14,15]
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm20 = [2,2,2,3,8,8,8,9]
+; AVX512F-FAST-NEXT: vmovdqa 32(%r9), %xmm3
; AVX512F-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,2,3,6,7,4,5,6,7,4,5,12,13,14,15]
-; AVX512F-FAST-NEXT: vpshufb %xmm0, %xmm3, %xmm7
-; AVX512F-FAST-NEXT: vpermi2q %zmm7, %zmm4, %zmm22
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; AVX512F-FAST-NEXT: vprold $16, %xmm2, %xmm2
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,2,3]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = [4,5,2,3,4,5,6,7,8,9,10,11,10,11,8,9]
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm4, %xmm2
+; AVX512F-FAST-NEXT: vmovdqa 32(%r8), %xmm5
+; AVX512F-FAST-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,2,3,6,7,4,5,6,7,4,5,12,13,14,15]
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm7, %xmm5
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm7, %xmm24
+; AVX512F-FAST-NEXT: vpermi2q %zmm5, %zmm2, %zmm20
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; AVX512F-FAST-NEXT: vprold $16, %xmm1, %xmm1
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm9 = [4,5,2,3,4,5,6,7,8,9,10,11,10,11,8,9]
+; AVX512F-FAST-NEXT: vpshufb %xmm9, %xmm2, %xmm1
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm26 = <u,0,u,1,8,8,9,u>
-; AVX512F-FAST-NEXT: vpermi2q %zmm2, %zmm1, %zmm26
+; AVX512F-FAST-NEXT: vpermi2q %zmm1, %zmm0, %zmm26
; AVX512F-FAST-NEXT: vmovdqa (%r9), %xmm2
-; AVX512F-FAST-NEXT: vmovdqa (%r8), %xmm9
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm9[0],xmm2[0],xmm9[1],xmm2[1],xmm9[2],xmm2[2],xmm9[3],xmm2[3]
-; AVX512F-FAST-NEXT: vpshufb %xmm0, %xmm4, %xmm0
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
+; AVX512F-FAST-NEXT: vmovdqa (%r8), %xmm1
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm0, %xmm3
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm27 = [0,0,0,1,8,9,9,11]
-; AVX512F-FAST-NEXT: vpermi2q %zmm4, %zmm0, %zmm27
+; AVX512F-FAST-NEXT: vpermi2q %zmm0, %zmm3, %zmm27
; AVX512F-FAST-NEXT: vprold $16, %ymm13, %ymm0
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm14[1,2,2,3,5,6,6,7]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0,1],ymm0[2],ymm4[3,4],ymm0[5],ymm4[6,7,8,9],ymm0[10],ymm4[11,12],ymm0[13],ymm4[14,15]
-; AVX512F-FAST-NEXT: vpbroadcastd {{.*#+}} ymm3 = [18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm13, %ymm4
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm13 = ymm14[0,0,2,1,4,4,6,5]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2],ymm13[3],ymm4[4,5],ymm13[6],ymm4[7,8,9,10],ymm13[11],ymm4[12,13],ymm13[14],ymm4[15]
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm3 = ymm14[1,2,2,3,5,6,6,7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0,1],ymm0[2],ymm3[3,4],ymm0[5],ymm3[6,7,8,9],ymm0[10],ymm3[11,12],ymm0[13],ymm3[14,15]
+; AVX512F-FAST-NEXT: vpbroadcastd {{.*#+}} ymm5 = [18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
+; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm13, %ymm3
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm7 = ymm14[0,0,2,1,4,4,6,5]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm7[3],ymm3[4,5],ymm7[6],ymm3[7,8,9,10],ymm7[11],ymm3[12,13],ymm7[14],ymm3[15]
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm28 = [2,2,3,3,10,9,11,10]
-; AVX512F-FAST-NEXT: vpermi2q %zmm0, %zmm4, %zmm28
-; AVX512F-FAST-NEXT: vmovdqa (%rdi), %xmm0
-; AVX512F-FAST-NEXT: vmovdqa (%rsi), %xmm14
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm14[4],xmm0[4],xmm14[5],xmm0[5],xmm14[6],xmm0[6],xmm14[7],xmm0[7]
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm4, %xmm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm19
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm9[4],xmm2[4],xmm9[5],xmm2[5],xmm9[6],xmm2[6],xmm9[7],xmm2[7]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm25, %zmm0, %zmm4
-; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm2, %xmm2
+; AVX512F-FAST-NEXT: vpermi2q %zmm0, %zmm3, %zmm28
+; AVX512F-FAST-NEXT: vmovdqa (%rdi), %xmm8
+; AVX512F-FAST-NEXT: vmovdqa (%rsi), %xmm0
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
+; AVX512F-FAST-NEXT: vpshufb %xmm9, %xmm3, %xmm3
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm17
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm25, %zmm0, %zmm2
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm30, %xmm9
+; AVX512F-FAST-NEXT: vpshufb %xmm9, %xmm1, %xmm1
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm25 = <0,0,1,1,12,13,u,15>
-; AVX512F-FAST-NEXT: vpermi2q %zmm4, %zmm2, %zmm25
+; AVX512F-FAST-NEXT: vpermi2q %zmm2, %zmm1, %zmm25
; AVX512F-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-FAST-NEXT: vpbroadcastd 8(%rax), %ymm2
-; AVX512F-FAST-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa (%rax), %ymm9
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [12,13,128,128,128,128,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128]
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm9, %ymm4
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm2, %zmm31
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm2
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm8[1,1,1,1,5,5,5,5]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm18
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm29, %ymm7
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm7[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm12[0,1,1,3,4,5,5,7]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0,1],ymm2[2],ymm4[3,4],ymm2[5],ymm4[6,7,8,9],ymm2[10],ymm4[11,12],ymm2[13],ymm4[14,15]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm17
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm15, %ymm2
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm3 = ymm6[0,0,2,1,4,4,6,5]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3],ymm2[4,5],ymm3[6],ymm2[7,8,9,10],ymm3[11],ymm2[12,13],ymm3[14],ymm2[15]
+; AVX512F-FAST-NEXT: vpbroadcastd 8(%rax), %ymm1
+; AVX512F-FAST-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa (%rax), %ymm3
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [12,13,128,128,128,128,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128]
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm7
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm1, %zmm30
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm29, %ymm1
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm6[1,1,1,1,5,5,5,5]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm6[2],ymm1[3,4],ymm6[5],ymm1[6,7,8,9],ymm6[10],ymm1[11,12],ymm6[13],ymm1[14,15]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm16
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm31, %ymm13
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm13[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm12[0,1,1,3,4,5,5,7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm6[0,1],ymm1[2],ymm6[3,4],ymm1[5],ymm6[6,7,8,9],ymm1[10],ymm6[11,12],ymm1[13],ymm6[14,15]
+; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm15, %ymm1
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm5 = ymm4[0,0,2,1,4,4,6,5]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm5[3],ymm1[4,5],ymm5[6],ymm1[7,8,9,10],ymm5[11],ymm1[12,13],ymm5[14],ymm1[15]
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm29 = <0,1,u,3,10,10,11,11>
-; AVX512F-FAST-NEXT: vpermi2q %zmm2, %zmm21, %zmm29
-; AVX512F-FAST-NEXT: vmovdqa 32(%rax), %ymm8
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm8, %ymm1
-; AVX512F-FAST-NEXT: vbroadcasti64x4 {{.*#+}} zmm21 = [0,13,4,0,0,14,5,0,0,13,4,0,0,14,5,0]
-; AVX512F-FAST-NEXT: # zmm21 = mem[0,1,2,3,0,1,2,3]
-; AVX512F-FAST-NEXT: vpermd %ymm8, %ymm21, %ymm2
+; AVX512F-FAST-NEXT: vpermi2q %zmm1, %zmm21, %zmm29
+; AVX512F-FAST-NEXT: vmovdqa 32(%rax), %ymm6
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm6, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [0,1,4,5,4,5,5,7]
+; AVX512F-FAST-NEXT: vpermd %ymm6, %ymm2, %ymm2
; AVX512F-FAST-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm2
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm16
-; AVX512F-FAST-NEXT: vmovdqa %ymm7, %ymm4
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm7[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm3 = ymm12[3,3,3,3,7,7,7,7]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5,6,7,8],ymm2[9],ymm3[10,11],ymm2[12],ymm3[13,14,15]
-; AVX512F-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm7 = [22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27,22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27]
-; AVX512F-FAST-NEXT: # ymm7 = mem[0,1,0,1]
-; AVX512F-FAST-NEXT: vpshufb %ymm7, %ymm4, %ymm13
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm14
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm13[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm5 = ymm12[3,3,3,3,7,7,7,7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm5[0],ymm2[1],ymm5[2,3],ymm2[4],ymm5[5,6,7,8],ymm2[9],ymm5[10,11],ymm2[12],ymm5[13,14,15]
+; AVX512F-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27,22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27]
+; AVX512F-FAST-NEXT: # ymm5 = mem[0,1,0,1]
+; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm13, %ymm13
; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm12 = ymm12[2,2,2,2,6,6,6,6]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm13[0,1],ymm12[2],ymm13[3,4],ymm12[5],ymm13[6,7,8,9],ymm12[10],ymm13[11,12],ymm12[13],ymm13[14,15]
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm30 = [2,2,2,3,8,10,10,11]
-; AVX512F-FAST-NEXT: vpermt2q %zmm3, %zmm30, %zmm12
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm15[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm13 = ymm6[3,3,3,3,7,7,7,7]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm13[2],ymm3[3,4],ymm13[5],ymm3[6,7,8,9],ymm13[10],ymm3[11,12],ymm13[13],ymm3[14,15]
-; AVX512F-FAST-NEXT: vprold $16, %ymm15, %ymm13
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm6[1,2,2,3,5,6,6,7]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm13[2],ymm6[3,4],ymm13[5],ymm6[6,7,8,9],ymm13[10],ymm6[11,12],ymm13[13],ymm6[14,15]
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm13 = [2,1,3,2,10,10,10,11]
-; AVX512F-FAST-NEXT: vpermt2q %zmm3, %zmm13, %zmm6
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm5[4],xmm10[4],xmm5[5],xmm10[5],xmm5[6],xmm10[6],xmm5[7],xmm10[7]
-; AVX512F-FAST-NEXT: vpbroadcastq {{.*#+}} xmm15 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
-; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm10, %xmm10
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[1,1,2,2]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0],xmm10[1],xmm5[2,3],xmm10[4],xmm5[5,6],xmm10[7]
-; AVX512F-FAST-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,2,3,3,4,5,6,7]
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,0,1,1,8,8,10,9]
-; AVX512F-FAST-NEXT: vpermt2q %zmm3, %zmm2, %zmm5
-; AVX512F-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX512F-FAST-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
-; AVX512F-FAST-NEXT: # xmm3 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm13 = ymm13[0,1],ymm12[2],ymm13[3,4],ymm12[5],ymm13[6,7,8,9],ymm12[10],ymm13[11,12],ymm12[13],ymm13[14,15]
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm21 = [2,2,2,3,8,10,10,11]
+; AVX512F-FAST-NEXT: vpermt2q %zmm2, %zmm21, %zmm13
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm15[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm12 = ymm4[3,3,3,3,7,7,7,7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm12[2],ymm2[3,4],ymm12[5],ymm2[6,7,8,9],ymm12[10],ymm2[11,12],ymm12[13],ymm2[14,15]
+; AVX512F-FAST-NEXT: vprold $16, %ymm15, %ymm12
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[1,2,2,3,5,6,6,7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm4[0,1],ymm12[2],ymm4[3,4],ymm12[5],ymm4[6,7,8,9],ymm12[10],ymm4[11,12],ymm12[13],ymm4[14,15]
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm31 = [2,1,3,2,10,10,10,11]
+; AVX512F-FAST-NEXT: vpermt2q %zmm2, %zmm31, %zmm12
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm18 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535]
+; AVX512F-FAST-NEXT: vpternlogq $226, %zmm22, %zmm18, %zmm13
+; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm13, %zmm12
+; AVX512F-FAST-NEXT: vmovdqa64 (%rax), %zmm15
+; AVX512F-FAST-NEXT: vbroadcasti64x4 {{.*#+}} zmm4 = [30,5,0,0,31,6,0,31,30,5,0,0,31,6,0,31]
+; AVX512F-FAST-NEXT: # zmm4 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-FAST-NEXT: vpermi2d %zmm15, %zmm6, %zmm4
+; AVX512F-FAST-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm12, %zmm4
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm11[4],xmm10[4],xmm11[5],xmm10[5],xmm11[6],xmm10[6],xmm11[7],xmm10[7]
+; AVX512F-FAST-NEXT: vpbroadcastq {{.*#+}} xmm12 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
+; AVX512F-FAST-NEXT: vpshufb %xmm12, %xmm10, %xmm10
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[1,1,2,2]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm11[0],xmm10[1],xmm11[2,3],xmm10[4],xmm11[5,6],xmm10[7]
+; AVX512F-FAST-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,2,3,3,4,5,6,7]
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm11 = [0,0,1,1,8,8,10,9]
+; AVX512F-FAST-NEXT: vpermt2q %zmm6, %zmm11, %zmm10
; AVX512F-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm1[0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
-; AVX512F-FAST-NEXT: vmovdqa %xmm11, %xmm1
-; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm3, %xmm3
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm4 = [0,1,1,3,8,8,9,9]
-; AVX512F-FAST-NEXT: vpermt2q %zmm3, %zmm4, %zmm10
-; AVX512F-FAST-NEXT: vprold $16, %xmm14, %xmm3
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} xmm11 = xmm0[1,1,2,3]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm11[0,1],xmm3[2],xmm11[3,4],xmm3[5],xmm11[6,7]
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1],xmm0[2],xmm14[2],xmm0[3],xmm14[3]
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm11
-; AVX512F-FAST-NEXT: vpermt2q %zmm3, %zmm2, %zmm11
-; AVX512F-FAST-NEXT: vmovdqa (%rcx), %xmm0
-; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm0, %xmm2
-; AVX512F-FAST-NEXT: vmovdqa (%rdx), %xmm3
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} xmm14 = xmm3[1,1,2,2]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm14[0],xmm2[1],xmm14[2,3],xmm2[4],xmm14[5,6],xmm2[7]
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm14 = xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,0,1,2,3,6,7,4,5,6,7,4,5]
-; AVX512F-FAST-NEXT: vpermt2q %zmm2, %zmm4, %zmm0
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535]
-; AVX512F-FAST-NEXT: vpternlogq $226, %zmm24, %zmm2, %zmm12
-; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm12, %zmm6
-; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX512F-FAST-NEXT: vpshufb %ymm7, %ymm12, %ymm3
-; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm1[2,2,2,2,6,6,6,6]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7,8,9],ymm4[10],ymm3[11,12],ymm4[13],ymm3[14,15]
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm4 = ymm12[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm7 = ymm1[0,1,1,3,4,5,5,7]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm7[0,1],ymm4[2],ymm7[3,4],ymm4[5],ymm7[6,7,8,9],ymm4[10],ymm7[11,12],ymm4[13],ymm7[14,15]
-; AVX512F-FAST-NEXT: vpermt2q %zmm3, %zmm13, %zmm4
-; AVX512F-FAST-NEXT: vmovdqa64 (%rax), %zmm3
-; AVX512F-FAST-NEXT: vbroadcasti64x4 {{.*#+}} zmm7 = [30,5,0,0,31,6,0,31,30,5,0,0,31,6,0,31]
-; AVX512F-FAST-NEXT: # zmm7 = mem[0,1,2,3,0,1,2,3]
-; AVX512F-FAST-NEXT: vpermi2d %zmm3, %zmm8, %zmm7
-; AVX512F-FAST-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm6, %zmm7
+; AVX512F-FAST-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm6 # 16-byte Folded Reload
+; AVX512F-FAST-NEXT: # xmm6 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm24, %xmm1
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm13 = xmm1[0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
+; AVX512F-FAST-NEXT: vpshufb %xmm9, %xmm6, %xmm6
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm22 = [0,1,1,3,8,8,9,9]
+; AVX512F-FAST-NEXT: vpermt2q %zmm6, %zmm22, %zmm13
+; AVX512F-FAST-NEXT: vprold $16, %xmm0, %xmm6
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} xmm2 = xmm8[1,1,2,3]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm6[2],xmm2[3,4],xmm6[5],xmm2[6,7]
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
+; AVX512F-FAST-NEXT: vpshufb %xmm9, %xmm0, %xmm0
+; AVX512F-FAST-NEXT: vpermt2q %zmm2, %zmm11, %zmm0
+; AVX512F-FAST-NEXT: vmovdqa (%rcx), %xmm2
+; AVX512F-FAST-NEXT: vpshufb %xmm12, %xmm2, %xmm6
+; AVX512F-FAST-NEXT: vmovdqa (%rdx), %xmm11
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} xmm12 = xmm11[1,1,2,2]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm12[0],xmm6[1],xmm12[2,3],xmm6[4],xmm12[5,6],xmm6[7]
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm12 = xmm11[4],xmm2[4],xmm11[5],xmm2[5],xmm11[6],xmm2[6],xmm11[7],xmm2[7]
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm11[0],xmm2[0],xmm11[1],xmm2[1],xmm11[2],xmm2[2],xmm11[3],xmm2[3]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[0,1,2,3,0,1,2,3,6,7,4,5,6,7,4,5]
+; AVX512F-FAST-NEXT: vpermt2q %zmm6, %zmm22, %zmm2
; AVX512F-FAST-NEXT: vpbroadcastd 36(%rax), %ymm6
-; AVX512F-FAST-NEXT: vpbroadcastd 40(%rax), %ymm8
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm6, %zmm6
-; AVX512F-FAST-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm6
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm8 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535]
-; AVX512F-FAST-NEXT: vpternlogq $226, %zmm26, %zmm8, %zmm5
-; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm5, %zmm6
-; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm11, %zmm0
-; AVX512F-FAST-NEXT: vpbroadcastd (%rax), %ymm5
-; AVX512F-FAST-NEXT: vpbroadcastd 4(%rax), %ymm10
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm10, %zmm5, %zmm5
-; AVX512F-FAST-NEXT: vpshuflw {{.*#+}} xmm10 = xmm14[0,2,3,3,4,5,6,7]
+; AVX512F-FAST-NEXT: vpbroadcastd 40(%rax), %ymm11
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm6, %zmm6
+; AVX512F-FAST-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm13, %zmm6
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm11 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535]
+; AVX512F-FAST-NEXT: vpternlogq $226, %zmm26, %zmm11, %zmm10
+; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm6
+; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm2
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm10, %ymm0
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm5 = ymm1[2,2,2,2,6,6,6,6]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm5[2],ymm0[3,4],ymm5[5],ymm0[6,7,8,9],ymm5[10],ymm0[11,12],ymm5[13],ymm0[14,15]
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm5 = ymm10[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm10 = ymm1[0,1,1,3,4,5,5,7]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm10[0,1],ymm5[2],ymm10[3,4],ymm5[5],ymm10[6,7,8,9],ymm5[10],ymm10[11,12],ymm5[13],ymm10[14,15]
+; AVX512F-FAST-NEXT: vpshuflw {{.*#+}} xmm10 = xmm12[0,2,3,3,4,5,6,7]
; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,0,2,1]
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm11 = ymm19[0,0,1,3]
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm12 = ymm18[2,2,2,3]
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm13 = ymm17[2,1,3,2]
-; AVX512F-FAST-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm27, %zmm5
-; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm5
-; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm15[10,11,8,9,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,u,u,u,u,26,27,24,25,u,u,u,u]
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm9 = ymm17[0,0,1,3]
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm16[2,2,2,3]
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm7 = ymm7[2,1,3,2]
+; AVX512F-FAST-NEXT: vpermt2q %zmm0, %zmm31, %zmm5
+; AVX512F-FAST-NEXT: vpbroadcastd (%rax), %ymm0
+; AVX512F-FAST-NEXT: vpbroadcastd 4(%rax), %ymm12
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm12, %zmm0, %zmm0
+; AVX512F-FAST-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm27, %zmm0
+; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm0
+; AVX512F-FAST-NEXT: vmovdqu (%rsp), %ymm13 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm13[10,11,8,9,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,u,u,u,u,26,27,24,25,u,u,u,u]
; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm14 = ymm1[2,2,2,2,6,6,6,6]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm14[0],ymm0[1],ymm14[2,3],ymm0[4],ymm14[5,6,7,8],ymm0[9],ymm14[10,11],ymm0[12],ymm14[13,14,15]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm14 = ymm15[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm15 = ymm1[1,1,1,1,5,5,5,5]
-; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3,4],ymm15[5],ymm14[6,7,8,9],ymm15[10],ymm14[11,12],ymm15[13],ymm14[14,15]
-; AVX512F-FAST-NEXT: vpermt2q %zmm0, %zmm30, %zmm14
-; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm14
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <6,u,u,u,7,u,u,7>
-; AVX512F-FAST-NEXT: vpermd %ymm9, %ymm0, %ymm0
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm3, %zmm3
-; AVX512F-FAST-NEXT: vpermd %zmm3, %zmm21, %zmm3
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm12 = ymm1[2,2,2,2,6,6,6,6]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm12[0],ymm2[1],ymm12[2,3],ymm2[4],ymm12[5,6,7,8],ymm2[9],ymm12[10,11],ymm2[12],ymm12[13,14,15]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm12 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
+; AVX512F-FAST-NEXT: vpshufd {{.*#+}} ymm13 = ymm1[1,1,1,1,5,5,5,5]
+; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm12[0,1],ymm13[2],ymm12[3,4],ymm13[5],ymm12[6,7,8,9],ymm13[10],ymm12[11,12],ymm13[13],ymm12[14,15]
+; AVX512F-FAST-NEXT: vpermt2q %zmm2, %zmm21, %zmm12
+; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm5, %zmm12
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <6,u,u,u,7,u,u,7>
+; AVX512F-FAST-NEXT: vpermd %ymm3, %ymm2, %ymm2
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm15, %zmm3
+; AVX512F-FAST-NEXT: vbroadcasti64x4 {{.*#+}} zmm5 = [0,13,4,0,0,14,5,0,0,13,4,0,0,14,5,0]
+; AVX512F-FAST-NEXT: # zmm5 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-FAST-NEXT: vpermd %zmm3, %zmm5, %zmm3
; AVX512F-FAST-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm28, %zmm3
-; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm14, %zmm3
-; AVX512F-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm10, %zmm4 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm11, %zmm9 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: vpternlogq $226, %zmm4, %zmm8, %zmm9
-; AVX512F-FAST-NEXT: vpternlogq $226, %zmm23, %zmm2, %zmm20
-; AVX512F-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm25, %zmm31
-; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm9, %zmm31
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm12, %zmm2, %zmm2
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm13, %zmm4, %zmm4
-; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm4
-; AVX512F-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm29, %zmm16
-; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm16
-; AVX512F-FAST-NEXT: vpbroadcastd 32(%rax), %ymm2
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
-; AVX512F-FAST-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm22, %zmm0
-; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm20, %zmm0
+; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm12, %zmm3
+; AVX512F-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm10, %zmm5 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm9 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: vpternlogq $226, %zmm5, %zmm11, %zmm9
+; AVX512F-FAST-NEXT: vpternlogq $226, %zmm23, %zmm18, %zmm19
+; AVX512F-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm25, %zmm30
+; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm9, %zmm30
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm5, %zmm5
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm8, %zmm7
+; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm5, %zmm7
+; AVX512F-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm29, %zmm14
+; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm7, %zmm14
+; AVX512F-FAST-NEXT: vpbroadcastd 32(%rax), %ymm5
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm2, %zmm2
+; AVX512F-FAST-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm20, %zmm2
+; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm19, %zmm2
; AVX512F-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT: vmovdqa64 %zmm3, 128(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm5, (%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm16, 320(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm0, (%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm14, 320(%rax)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm6, 256(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm0, 192(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm7, 384(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm31, 64(%rax)
-; AVX512F-FAST-NEXT: addq $248, %rsp
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm2, 192(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm4, 384(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm30, 64(%rax)
+; AVX512F-FAST-NEXT: addq $232, %rsp
; AVX512F-FAST-NEXT: vzeroupper
; AVX512F-FAST-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-8.ll
index c02348af872917..bc77e91a2bd404 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-8.ll
@@ -3480,200 +3480,203 @@ define void @store_i16_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX512F-FAST-LABEL: store_i16_stride8_vf32:
; AVX512F-FAST: # %bb.0:
-; AVX512F-FAST-NEXT: subq $472, %rsp # imm = 0x1D8
+; AVX512F-FAST-NEXT: subq $520, %rsp # imm = 0x208
; AVX512F-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT: movq {{[0-9]+}}(%rsp), %r10
; AVX512F-FAST-NEXT: vmovdqa (%r10), %xmm1
; AVX512F-FAST-NEXT: vmovdqa (%rax), %xmm2
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm2, %xmm26
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm1, %xmm27
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm2, %xmm25
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm1, %xmm26
; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa (%r9), %xmm0
-; AVX512F-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-FAST-NEXT: vmovdqa (%r9), %xmm1
; AVX512F-FAST-NEXT: vmovdqa (%r8), %xmm2
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm2, %xmm31
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm2, %xmm27
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm1, %xmm28
; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vmovdqa (%rcx), %xmm1
; AVX512F-FAST-NEXT: vmovdqa (%rdx), %xmm2
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm2, %xmm20
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm1, %xmm21
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm2, %xmm31
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm1, %xmm19
; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vmovdqa (%rsi), %xmm1
; AVX512F-FAST-NEXT: vmovdqa (%rdi), %xmm2
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm2, %xmm18
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm1, %xmm19
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm0, %zmm0, %zmm0
-; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm2, %xmm17
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm1, %xmm18
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm0, %zmm0, %zmm20
; AVX512F-FAST-NEXT: vmovdqa 32(%r10), %ymm0
; AVX512F-FAST-NEXT: vmovdqa 32(%rax), %ymm1
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm24
-; AVX512F-FAST-NEXT: vmovdqa 32(%r9), %ymm2
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm23
+; AVX512F-FAST-NEXT: vmovdqa 32(%r9), %ymm3
; AVX512F-FAST-NEXT: vmovdqa 32(%r8), %ymm4
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[8],ymm2[8],ymm4[9],ymm2[9],ymm4[10],ymm2[10],ymm4[11],ymm2[11]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm3, %zmm25
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[8],ymm3[8],ymm4[9],ymm3[9],ymm4[10],ymm3[10],ymm4[11],ymm3[11]
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm24
; AVX512F-FAST-NEXT: vmovdqa 32(%rcx), %ymm5
; AVX512F-FAST-NEXT: vmovdqa 32(%rdx), %ymm6
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm6[0],ymm5[0],ymm6[1],ymm5[1],ymm6[2],ymm5[2],ymm6[3],ymm5[3],ymm6[8],ymm5[8],ymm6[9],ymm5[9],ymm6[10],ymm5[10],ymm6[11],ymm5[11]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm3, %zmm22
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm6[0],ymm5[0],ymm6[1],ymm5[1],ymm6[2],ymm5[2],ymm6[3],ymm5[3],ymm6[8],ymm5[8],ymm6[9],ymm5[9],ymm6[10],ymm5[10],ymm6[11],ymm5[11]
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm21
; AVX512F-FAST-NEXT: vmovdqa 32(%rsi), %ymm7
; AVX512F-FAST-NEXT: vmovdqa 32(%rdi), %ymm8
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm8[0],ymm7[0],ymm8[1],ymm7[1],ymm8[2],ymm7[2],ymm8[3],ymm7[3],ymm8[8],ymm7[8],ymm8[9],ymm7[9],ymm8[10],ymm7[10],ymm8[11],ymm7[11]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm3, %zmm23
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm8[0],ymm7[0],ymm8[1],ymm7[1],ymm8[2],ymm7[2],ymm8[3],ymm7[3],ymm8[8],ymm7[8],ymm8[9],ymm7[9],ymm8[10],ymm7[10],ymm8[11],ymm7[11]
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm22
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm1 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
; AVX512F-FAST-NEXT: vmovdqa 32(%r10), %xmm0
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm1, %zmm16
-; AVX512F-FAST-NEXT: vmovdqa 32(%rax), %xmm3
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm4[4],ymm2[4],ymm4[5],ymm2[5],ymm4[6],ymm2[6],ymm4[7],ymm2[7],ymm4[12],ymm2[12],ymm4[13],ymm2[13],ymm4[14],ymm2[14],ymm4[15],ymm2[15]
+; AVX512F-FAST-NEXT: vmovdqa 32(%rax), %xmm2
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm1 = ymm4[4],ymm3[4],ymm4[5],ymm3[5],ymm4[6],ymm3[6],ymm4[7],ymm3[7],ymm4[12],ymm3[12],ymm4[13],ymm3[13],ymm4[14],ymm3[14],ymm4[15],ymm3[15]
+; AVX512F-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vmovdqa 32(%r9), %xmm1
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm17
-; AVX512F-FAST-NEXT: vmovdqa 32(%r8), %xmm4
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm6[4],ymm5[4],ymm6[5],ymm5[5],ymm6[6],ymm5[6],ymm6[7],ymm5[7],ymm6[12],ymm5[12],ymm6[13],ymm5[13],ymm6[14],ymm5[14],ymm6[15],ymm5[15]
-; AVX512F-FAST-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa 32(%rcx), %xmm2
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm5 = ymm8[4],ymm7[4],ymm8[5],ymm7[5],ymm8[6],ymm7[6],ymm8[7],ymm7[7],ymm8[12],ymm7[12],ymm8[13],ymm7[13],ymm8[14],ymm7[14],ymm8[15],ymm7[15]
-; AVX512F-FAST-NEXT: vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
-; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm15 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm6[4],ymm5[4],ymm6[5],ymm5[5],ymm6[6],ymm5[6],ymm6[7],ymm5[7],ymm6[12],ymm5[12],ymm6[13],ymm5[13],ymm6[14],ymm5[14],ymm6[15],ymm5[15]
+; AVX512F-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-FAST-NEXT: vmovdqa 32(%r8), %xmm3
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm8[4],ymm7[4],ymm8[5],ymm7[5],ymm8[6],ymm7[6],ymm8[7],ymm7[7],ymm8[12],ymm7[12],ymm8[13],ymm7[13],ymm8[14],ymm7[14],ymm8[15],ymm7[15]
+; AVX512F-FAST-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; AVX512F-FAST-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa 32(%rdx), %xmm0
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; AVX512F-FAST-NEXT: vmovdqa 32(%rcx), %xmm0
+; AVX512F-FAST-NEXT: vmovdqa 32(%rdx), %xmm1
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vmovdqa 32(%rsi), %xmm0
; AVX512F-FAST-NEXT: vmovdqa 32(%rdi), %xmm1
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm14 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm15 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX512F-FAST-NEXT: vmovdqa (%r10), %ymm0
; AVX512F-FAST-NEXT: vmovdqa (%rax), %ymm1
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm12 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm10 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm13 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm11 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
; AVX512F-FAST-NEXT: vmovdqa (%r9), %ymm0
; AVX512F-FAST-NEXT: vmovdqa (%r8), %ymm1
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm6 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm7 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm7 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm8 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
; AVX512F-FAST-NEXT: vmovdqa (%rcx), %ymm0
; AVX512F-FAST-NEXT: vmovdqa (%rdx), %ymm1
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm5 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm6 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm5 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
; AVX512F-FAST-NEXT: vmovdqa (%rsi), %ymm0
-; AVX512F-FAST-NEXT: vmovdqa (%rdi), %ymm2
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11]
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm26, %xmm0
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm27, %xmm1
+; AVX512F-FAST-NEXT: vmovdqa (%rdi), %ymm1
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm25, %xmm0
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm26, %xmm1
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm27, %xmm0
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm28, %xmm1
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm25
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm26
; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm27
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm11, %zmm11, %zmm30
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm15, %zmm15, %zmm15
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm9, %zmm9, %zmm28
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm29
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm0, %zmm0, %zmm28
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm12, %zmm12, %zmm12
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm9, %zmm9, %zmm29
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm10, %zmm10, %zmm30
; AVX512F-FAST-NEXT: vmovdqa64 %xmm31, %xmm0
-; AVX512F-FAST-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm9 # 16-byte Folded Reload
-; AVX512F-FAST-NEXT: # xmm9 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm20, %xmm0
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm21, %xmm8
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm18, %xmm0
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm19, %xmm11
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm19, %xmm9
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm17, %xmm0
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm18, %xmm10
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm0[0],xmm10[0],xmm0[1],xmm10[1],xmm0[2],xmm10[2],xmm0[3],xmm10[3]
; AVX512F-FAST-NEXT: movw $-30584, %ax # imm = 0x8888
; AVX512F-FAST-NEXT: kmovw %eax, %k1
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm31 = [0,1,4,5,4,5,5,7,10,9,14,13,14,13,15,15]
-; AVX512F-FAST-NEXT: vpermd %zmm25, %zmm31, %zmm25
+; AVX512F-FAST-NEXT: vpermd %zmm24, %zmm31, %zmm24
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,0,4,4,4,4,6,5,8,10,12,14,12,14,14,15]
-; AVX512F-FAST-NEXT: vpermd %zmm24, %zmm0, %zmm25 {%k1}
-; AVX512F-FAST-NEXT: vpermd %zmm17, %zmm31, %zmm17
-; AVX512F-FAST-NEXT: vpermd %zmm16, %zmm0, %zmm17 {%k1}
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm13, %zmm13, %zmm16
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm13, %zmm13, %zmm24
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm13, %zmm13, %zmm13
+; AVX512F-FAST-NEXT: vpermd %zmm23, %zmm0, %zmm24 {%k1}
+; AVX512F-FAST-NEXT: vpermd %zmm25, %zmm31, %zmm23
+; AVX512F-FAST-NEXT: vpermd %zmm16, %zmm0, %zmm23 {%k1}
+; AVX512F-FAST-NEXT: vmovdqu64 (%rsp), %zmm14 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm14, %zmm14, %zmm16
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm14, %zmm14, %zmm25
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm14, %zmm14, %zmm14
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm12, %zmm12, %zmm12
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm15, %zmm15, %zmm15
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm13, %zmm13, %zmm13
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm7, %zmm7
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm6, %zmm6
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm5, %zmm5
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm3, %zmm3
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm10, %zmm10, %zmm10
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm7, %zmm7
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm11, %zmm11
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm8, %zmm8
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm5, %zmm5
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm4, %zmm4
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm2
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm2, %zmm2, %zmm2
; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm1, %zmm1, %zmm1
; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm9, %zmm9, %zmm9
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm8
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm11, %zmm11, %zmm11
-; AVX512F-FAST-NEXT: vpermd %zmm6, %zmm31, %zmm6
-; AVX512F-FAST-NEXT: vpermd %zmm12, %zmm0, %zmm6 {%k1}
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm10, %zmm10, %zmm10
; AVX512F-FAST-NEXT: vpermd %zmm7, %zmm31, %zmm7
+; AVX512F-FAST-NEXT: vpermd %zmm13, %zmm0, %zmm7 {%k1}
+; AVX512F-FAST-NEXT: vpermd %zmm8, %zmm31, %zmm8
; AVX512F-FAST-NEXT: movw $8738, %ax # imm = 0x2222
; AVX512F-FAST-NEXT: kmovw %eax, %k2
-; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm0, %zmm7 {%k1}
+; AVX512F-FAST-NEXT: vpermd %zmm11, %zmm0, %zmm8 {%k1}
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [4,5,1,3,5,7,5,7,14,13,11,11,15,15,15,15]
-; AVX512F-FAST-NEXT: vpermd %zmm23, %zmm0, %zmm10
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm12 = [4,4,2,1,6,5,6,5,12,14,10,11,14,15,14,15]
-; AVX512F-FAST-NEXT: vpermd %zmm22, %zmm12, %zmm10 {%k2}
-; AVX512F-FAST-NEXT: vpermd %zmm27, %zmm0, %zmm22
-; AVX512F-FAST-NEXT: vpermd %zmm26, %zmm12, %zmm22 {%k2}
+; AVX512F-FAST-NEXT: vpermd %zmm22, %zmm0, %zmm11
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm13 = [4,4,2,1,6,5,6,5,12,14,10,11,14,15,14,15]
+; AVX512F-FAST-NEXT: vpermd %zmm21, %zmm13, %zmm11 {%k2}
+; AVX512F-FAST-NEXT: vpermd %zmm27, %zmm0, %zmm21
+; AVX512F-FAST-NEXT: vpermd %zmm26, %zmm13, %zmm21 {%k2}
; AVX512F-FAST-NEXT: vpermd %zmm3, %zmm0, %zmm3
-; AVX512F-FAST-NEXT: vpermd %zmm5, %zmm12, %zmm3 {%k2}
-; AVX512F-FAST-NEXT: vpermd %zmm2, %zmm0, %zmm0
-; AVX512F-FAST-NEXT: vpermd %zmm4, %zmm12, %zmm0 {%k2}
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm2 = <0,u,0,u,u,u,1,u,10,10,10,10,u,u,11,11>
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm4 # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm5 = <0,0,0,0,u,u,1,1,10,10,10,10,u,u,11,11>
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm5, %zmm4 {%k1} # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vpermd %zmm15, %zmm2, %zmm12
-; AVX512F-FAST-NEXT: vpermd %zmm30, %zmm5, %zmm12 {%k1}
-; AVX512F-FAST-NEXT: vpermd %zmm24, %zmm2, %zmm15
-; AVX512F-FAST-NEXT: vpermd %zmm16, %zmm5, %zmm15 {%k1}
-; AVX512F-FAST-NEXT: vpermd %zmm9, %zmm2, %zmm2
-; AVX512F-FAST-NEXT: vpermd %zmm1, %zmm5, %zmm2 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = <0,u,1,u,1,u,u,u,10,u,11,u,11,u,u,u>
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm1 # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm5 = <0,0,1,1,1,1,u,u,10,10,11,11,11,11,u,u>
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm5, %zmm1 {%k2} # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm9 = <0,u,1,u,1,u,u,u,10,10,11,11,11,11,u,u>
-; AVX512F-FAST-NEXT: vpermd %zmm29, %zmm9, %zmm16
-; AVX512F-FAST-NEXT: vpermd %zmm28, %zmm5, %zmm16 {%k2}
-; AVX512F-FAST-NEXT: vpermd %zmm14, %zmm9, %zmm14
-; AVX512F-FAST-NEXT: vpermd %zmm13, %zmm5, %zmm14 {%k2}
-; AVX512F-FAST-NEXT: vpermd %zmm11, %zmm9, %zmm9
-; AVX512F-FAST-NEXT: vpermd %zmm8, %zmm5, %zmm9 {%k2}
+; AVX512F-FAST-NEXT: vpermd %zmm6, %zmm13, %zmm3 {%k2}
+; AVX512F-FAST-NEXT: vpermd %zmm4, %zmm0, %zmm0
+; AVX512F-FAST-NEXT: vpermd %zmm5, %zmm13, %zmm0 {%k2}
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm4 = <0,0,0,0,u,u,1,1,10,10,10,10,u,u,11,11>
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm5 # 64-byte Folded Reload
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm5 {%k1} # 64-byte Folded Reload
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm6 = <0,u,0,u,u,u,1,u,10,10,10,10,u,u,11,11>
+; AVX512F-FAST-NEXT: vpermd %zmm12, %zmm6, %zmm6
+; AVX512F-FAST-NEXT: vpermd %zmm28, %zmm4, %zmm6 {%k1}
+; AVX512F-FAST-NEXT: vpermd %zmm25, %zmm4, %zmm12
+; AVX512F-FAST-NEXT: vpermd %zmm16, %zmm4, %zmm12 {%k1}
+; AVX512F-FAST-NEXT: vpermd %zmm1, %zmm4, %zmm1
+; AVX512F-FAST-NEXT: vpermd %zmm2, %zmm4, %zmm1 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm2 = <0,0,1,1,1,1,u,u,10,u,11,u,11,u,u,u>
+; AVX512F-FAST-NEXT: vpermd %zmm20, %zmm2, %zmm2
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm4 = <0,0,1,1,1,1,u,u,10,10,11,11,11,11,u,u>
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm2 {%k2} # 64-byte Folded Reload
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm13 = <0,u,1,u,1,u,u,u,10,10,11,11,11,11,u,u>
+; AVX512F-FAST-NEXT: vpermd %zmm30, %zmm13, %zmm13
+; AVX512F-FAST-NEXT: vpermd %zmm29, %zmm4, %zmm13 {%k2}
+; AVX512F-FAST-NEXT: vpermd %zmm15, %zmm4, %zmm15
+; AVX512F-FAST-NEXT: vpermd %zmm14, %zmm4, %zmm15 {%k2}
+; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm4, %zmm10
+; AVX512F-FAST-NEXT: vpermd %zmm9, %zmm4, %zmm10 {%k2}
; AVX512F-FAST-NEXT: movb $-86, %al
; AVX512F-FAST-NEXT: kmovw %eax, %k1
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm4, %zmm1 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm25, %zmm10 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm17, %zmm22 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm12, %zmm16 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm15, %zmm14 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm6, %zmm3 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm7, %zmm0 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm2, %zmm9 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm5, %zmm2 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm24, %zmm11 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm23, %zmm21 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm6, %zmm13 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm12, %zmm15 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm7, %zmm3 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm8, %zmm0 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm1, %zmm10 {%k1}
; AVX512F-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm9, (%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm10, (%rax)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm0, 192(%rax)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm3, 128(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm14, 320(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm16, 256(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm22, 448(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm10, 384(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm1, 64(%rax)
-; AVX512F-FAST-NEXT: addq $472, %rsp # imm = 0x1D8
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm15, 320(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm13, 256(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm21, 448(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm11, 384(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm2, 64(%rax)
+; AVX512F-FAST-NEXT: addq $520, %rsp # imm = 0x208
; AVX512F-FAST-NEXT: vzeroupper
; AVX512F-FAST-NEXT: retq
;
@@ -7248,11 +7251,11 @@ define void @store_i16_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-FAST-NEXT: vmovdqa 32(%r10), %ymm0
; AVX512F-FAST-NEXT: vmovdqa 32(%rax), %ymm2
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm3, %zmm27
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm3, %zmm17
; AVX512F-FAST-NEXT: vmovdqa 32(%r9), %ymm3
; AVX512F-FAST-NEXT: vmovdqa 32(%r8), %ymm4
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm5 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[8],ymm3[8],ymm4[9],ymm3[9],ymm4[10],ymm3[10],ymm4[11],ymm3[11]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm5, %zmm17
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm5, %zmm15
; AVX512F-FAST-NEXT: vmovdqa 32(%rcx), %ymm5
; AVX512F-FAST-NEXT: vmovdqa 32(%rdx), %ymm6
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm7 = ymm6[0],ymm5[0],ymm6[1],ymm5[1],ymm6[2],ymm5[2],ymm6[3],ymm5[3],ymm6[8],ymm5[8],ymm6[9],ymm5[9],ymm6[10],ymm5[10],ymm6[11],ymm5[11]
@@ -7265,63 +7268,63 @@ define void @store_i16_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm4[4],ymm3[4],ymm4[5],ymm3[5],ymm4[6],ymm3[6],ymm4[7],ymm3[7],ymm4[12],ymm3[12],ymm4[13],ymm3[13],ymm4[14],ymm3[14],ymm4[15],ymm3[15]
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm6[4],ymm5[4],ymm6[5],ymm5[5],ymm6[6],ymm5[6],ymm6[7],ymm5[7],ymm6[12],ymm5[12],ymm6[13],ymm5[13],ymm6[14],ymm5[14],ymm6[15],ymm5[15]
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm8[4],ymm7[4],ymm8[5],ymm7[5],ymm8[6],ymm7[6],ymm8[7],ymm7[7],ymm8[12],ymm7[12],ymm8[13],ymm7[13],ymm8[14],ymm7[14],ymm8[15],ymm7[15]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm9, %zmm15
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm3, %zmm13
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm9, %zmm13
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm3, %zmm7
; AVX512F-FAST-NEXT: vmovdqa 32(%r10), %xmm3
; AVX512F-FAST-NEXT: vmovdqa 32(%rax), %xmm4
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm21
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm22
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm9
; AVX512F-FAST-NEXT: vmovdqa 32(%r9), %xmm0
; AVX512F-FAST-NEXT: vmovdqa 32(%r8), %xmm2
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm5, %zmm26
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm4, %zmm4, %zmm23
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm5, %zmm27
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm4, %zmm4, %zmm26
; AVX512F-FAST-NEXT: vmovdqa 32(%rcx), %xmm2
; AVX512F-FAST-NEXT: vmovdqa 32(%rdx), %xmm4
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
; AVX512F-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7]
; AVX512F-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm3, %zmm3, %zmm22
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm0, %zmm0, %zmm20
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm3, %zmm3, %zmm23
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm0, %zmm0, %zmm21
; AVX512F-FAST-NEXT: vmovdqa 32(%rsi), %xmm0
; AVX512F-FAST-NEXT: vmovdqa 32(%rdi), %xmm2
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX512F-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa (%r10), %ymm6
+; AVX512F-FAST-NEXT: vmovdqa (%r10), %ymm5
; AVX512F-FAST-NEXT: vmovdqa (%rax), %ymm3
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm6[0],ymm3[1],ymm6[1],ymm3[2],ymm6[2],ymm3[3],ymm6[3],ymm3[8],ymm6[8],ymm3[9],ymm6[9],ymm3[10],ymm6[10],ymm3[11],ymm6[11]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm7
-; AVX512F-FAST-NEXT: vmovdqa (%r9), %ymm5
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm5[0],ymm3[1],ymm5[1],ymm3[2],ymm5[2],ymm3[3],ymm5[3],ymm3[8],ymm5[8],ymm3[9],ymm5[9],ymm3[10],ymm5[10],ymm3[11],ymm5[11]
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm6
+; AVX512F-FAST-NEXT: vmovdqa (%r9), %ymm8
; AVX512F-FAST-NEXT: vmovdqa (%r8), %ymm10
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm4 = ymm10[0],ymm5[0],ymm10[1],ymm5[1],ymm10[2],ymm5[2],ymm10[3],ymm5[3],ymm10[8],ymm5[8],ymm10[9],ymm5[9],ymm10[10],ymm5[10],ymm10[11],ymm5[11]
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm4 = ymm10[0],ymm8[0],ymm10[1],ymm8[1],ymm10[2],ymm8[2],ymm10[3],ymm8[3],ymm10[8],ymm8[8],ymm10[9],ymm8[9],ymm10[10],ymm8[10],ymm10[11],ymm8[11]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm4, %zmm19
; AVX512F-FAST-NEXT: vmovdqa (%rcx), %ymm11
; AVX512F-FAST-NEXT: vmovdqa (%rdx), %ymm14
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm4 = ymm14[0],ymm11[0],ymm14[1],ymm11[1],ymm14[2],ymm11[2],ymm14[3],ymm11[3],ymm14[8],ymm11[8],ymm14[9],ymm11[9],ymm14[10],ymm11[10],ymm14[11],ymm11[11]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm4, %zmm16
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm4, %zmm20
; AVX512F-FAST-NEXT: vmovdqa (%rsi), %ymm1
; AVX512F-FAST-NEXT: vmovdqa (%rdi), %ymm0
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} ymm4 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm4, %zmm4
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm3[4],ymm6[4],ymm3[5],ymm6[5],ymm3[6],ymm6[6],ymm3[7],ymm6[7],ymm3[12],ymm6[12],ymm3[13],ymm6[13],ymm3[14],ymm6[14],ymm3[15],ymm6[15]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm3, %zmm6
-; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm10[4],ymm5[4],ymm10[5],ymm5[5],ymm10[6],ymm5[6],ymm10[7],ymm5[7],ymm10[12],ymm5[12],ymm10[13],ymm5[13],ymm10[14],ymm5[14],ymm10[15],ymm5[15]
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm4, %zmm16
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm3[4],ymm5[4],ymm3[5],ymm5[5],ymm3[6],ymm5[6],ymm3[7],ymm5[7],ymm3[12],ymm5[12],ymm3[13],ymm5[13],ymm3[14],ymm5[14],ymm3[15],ymm5[15]
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm3, %zmm4
+; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm10[4],ymm8[4],ymm10[5],ymm8[5],ymm10[6],ymm8[6],ymm10[7],ymm8[7],ymm10[12],ymm8[12],ymm10[13],ymm8[13],ymm10[14],ymm8[14],ymm10[15],ymm8[15]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm3, %zmm2
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm14[4],ymm11[4],ymm14[5],ymm11[5],ymm14[6],ymm11[6],ymm14[7],ymm11[7],ymm14[12],ymm11[12],ymm14[13],ymm11[13],ymm14[14],ymm11[14],ymm14[15],ymm11[15]
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
; AVX512F-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3]
; AVX512F-FAST-NEXT: vmovdqa (%rsp), %xmm5 # 16-byte Reload
-; AVX512F-FAST-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX512F-FAST-NEXT: # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1],xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX512F-FAST-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm8 # 16-byte Folded Reload
+; AVX512F-FAST-NEXT: # xmm8 = xmm5[0],mem[0],xmm5[1],mem[1],xmm5[2],mem[2],xmm5[3],mem[3]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm3, %zmm10
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm8
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm5
; AVX512F-FAST-NEXT: vmovdqa (%rcx), %xmm0
; AVX512F-FAST-NEXT: vmovdqa (%rdx), %xmm3
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm11 = xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
@@ -7329,7 +7332,7 @@ define void @store_i16_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm1, %zmm1, %zmm14
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm5, %zmm12
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm12
; AVX512F-FAST-NEXT: vmovdqa (%rsi), %xmm0
; AVX512F-FAST-NEXT: vmovdqa (%rdi), %xmm1
; AVX512F-FAST-NEXT: vpunpckhwd {{.*#+}} xmm11 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
@@ -7348,16 +7351,16 @@ define void @store_i16_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-FAST-NEXT: vpermd %zmm25, %zmm1, %zmm24 {%k2}
; AVX512F-FAST-NEXT: vpermd %zmm18, %zmm0, %zmm25
; AVX512F-FAST-NEXT: vpermd %zmm30, %zmm1, %zmm25 {%k2}
-; AVX512F-FAST-NEXT: vpermd %zmm17, %zmm0, %zmm5
-; AVX512F-FAST-NEXT: vpermd %zmm27, %zmm1, %zmm5 {%k2}
-; AVX512F-FAST-NEXT: vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vpermd %zmm13, %zmm0, %zmm27
-; AVX512F-FAST-NEXT: vpermd %zmm15, %zmm1, %zmm27 {%k2}
-; AVX512F-FAST-NEXT: vpermd %zmm19, %zmm0, %zmm5
-; AVX512F-FAST-NEXT: vpermd %zmm7, %zmm1, %zmm5 {%k2}
-; AVX512F-FAST-NEXT: vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-FAST-NEXT: vpermd %zmm15, %zmm0, %zmm8
+; AVX512F-FAST-NEXT: vpermd %zmm17, %zmm1, %zmm8 {%k2}
+; AVX512F-FAST-NEXT: vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-FAST-NEXT: vpermd %zmm7, %zmm0, %zmm30
+; AVX512F-FAST-NEXT: vpermd %zmm13, %zmm1, %zmm30 {%k2}
+; AVX512F-FAST-NEXT: vpermd %zmm19, %zmm0, %zmm7
+; AVX512F-FAST-NEXT: vpermd %zmm6, %zmm1, %zmm7 {%k2}
+; AVX512F-FAST-NEXT: vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vpermd %zmm2, %zmm0, %zmm19
-; AVX512F-FAST-NEXT: vpermd %zmm6, %zmm1, %zmm19 {%k2}
+; AVX512F-FAST-NEXT: vpermd %zmm4, %zmm1, %zmm19 {%k2}
; AVX512F-FAST-NEXT: movw $8738, %ax # imm = 0x2222
; AVX512F-FAST-NEXT: kmovw %eax, %k1
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [4,5,1,3,5,7,5,7,14,13,11,11,15,15,15,15]
@@ -7373,114 +7376,114 @@ define void @store_i16_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-FAST-NEXT: vpermd %zmm28, %zmm0, %zmm17
; AVX512F-FAST-NEXT: vpermd %zmm29, %zmm1, %zmm17 {%k1}
; AVX512F-FAST-NEXT: vpermd %zmm9, %zmm0, %zmm9
-; AVX512F-FAST-NEXT: vpermd %zmm21, %zmm1, %zmm9 {%k1}
-; AVX512F-FAST-NEXT: vpermd %zmm4, %zmm0, %zmm4
-; AVX512F-FAST-NEXT: vpermd %zmm16, %zmm1, %zmm4 {%k1}
-; AVX512F-FAST-NEXT: vpermd %zmm8, %zmm0, %zmm8
-; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm1, %zmm8 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = <0,u,0,u,u,u,1,u,10,10,10,10,u,u,11,11>
+; AVX512F-FAST-NEXT: vpermd %zmm22, %zmm1, %zmm9 {%k1}
+; AVX512F-FAST-NEXT: vpermd %zmm16, %zmm0, %zmm4
+; AVX512F-FAST-NEXT: vpermd %zmm20, %zmm1, %zmm4 {%k1}
+; AVX512F-FAST-NEXT: vpermd %zmm5, %zmm0, %zmm5
+; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm1, %zmm5 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = <0,0,0,0,u,u,1,1,10,10,10,10,u,u,11,11>
; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm10 = <0,0,0,0,u,u,1,1,10,10,10,10,u,u,11,11>
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm10, %zmm1 {%k2} # 64-byte Folded Reload
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 {%k2} # 64-byte Folded Reload
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm10 = <0,u,0,u,u,u,1,u,10,10,10,10,u,u,11,11>
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm10, %zmm10 # 64-byte Folded Reload
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm10 {%k2} # 64-byte Folded Reload
; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm16 # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm10, %zmm16 {%k2} # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm21 # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm10, %zmm21 {%k2} # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm28 # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm10, %zmm28 {%k2} # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm29 # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm10, %zmm29 {%k2} # 64-byte Folded Reload
-; AVX512F-FAST-NEXT: vpermd %zmm23, %zmm0, %zmm23
-; AVX512F-FAST-NEXT: vpermd %zmm26, %zmm10, %zmm23 {%k2}
-; AVX512F-FAST-NEXT: vpermd %zmm20, %zmm0, %zmm20
-; AVX512F-FAST-NEXT: vpermd %zmm22, %zmm10, %zmm20 {%k2}
-; AVX512F-FAST-NEXT: vpermd %zmm12, %zmm0, %zmm0
-; AVX512F-FAST-NEXT: vpermd %zmm14, %zmm10, %zmm0 {%k2}
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm11, %zmm11, %zmm10
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm11 = <0,u,1,u,1,u,u,u,10,u,11,u,11,u,u,u>
-; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm11, %zmm12
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm2, %zmm2, %zmm10
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm16 {%k2} # 64-byte Folded Reload
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm20 # 64-byte Folded Reload
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm20 {%k2} # 64-byte Folded Reload
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm22 # 64-byte Folded Reload
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm22 {%k2} # 64-byte Folded Reload
+; AVX512F-FAST-NEXT: vpermd %zmm26, %zmm0, %zmm26
+; AVX512F-FAST-NEXT: vpermd %zmm27, %zmm0, %zmm26 {%k2}
+; AVX512F-FAST-NEXT: vpermd %zmm21, %zmm0, %zmm21
+; AVX512F-FAST-NEXT: vpermd %zmm23, %zmm0, %zmm21 {%k2}
+; AVX512F-FAST-NEXT: vpermd %zmm12, %zmm0, %zmm23
+; AVX512F-FAST-NEXT: vpermd %zmm14, %zmm0, %zmm23 {%k2}
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm11, %zmm11, %zmm0
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm11 = <0,0,1,1,1,1,u,u,10,u,11,u,11,u,u,u>
+; AVX512F-FAST-NEXT: vpermd %zmm0, %zmm11, %zmm12
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm11 = <0,0,1,1,1,1,u,u,10,10,11,11,11,11,u,u>
-; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm11, %zmm12 {%k1}
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm5, %zmm10
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm5, %zmm14
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm22 = <0,u,1,u,1,u,u,u,10,10,11,11,11,11,u,u>
-; AVX512F-FAST-NEXT: vpermd %zmm14, %zmm22, %zmm14
-; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm11, %zmm14 {%k1}
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm5, %zmm10
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm5, %zmm26
-; AVX512F-FAST-NEXT: vpermd %zmm26, %zmm22, %zmm26
-; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm11, %zmm26 {%k1}
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm5, %zmm10
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm5, %zmm30
-; AVX512F-FAST-NEXT: vpermd %zmm30, %zmm22, %zmm30
-; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm11, %zmm30 {%k1}
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm5, %zmm10
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm5, %zmm31
-; AVX512F-FAST-NEXT: vpermd %zmm31, %zmm22, %zmm31
-; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm11, %zmm31 {%k1}
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm5, %zmm5, %zmm10
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm2, %zmm2, %zmm5
-; AVX512F-FAST-NEXT: vpermd %zmm5, %zmm22, %zmm5
-; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm11, %zmm5 {%k1}
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm10 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm10, %zmm10, %zmm10
+; AVX512F-FAST-NEXT: vpermd %zmm0, %zmm11, %zmm12 {%k1}
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm0, %zmm0, %zmm0
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm14 = <0,u,1,u,1,u,u,u,10,10,11,11,11,11,u,u>
+; AVX512F-FAST-NEXT: vpermd %zmm0, %zmm14, %zmm0
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm14
+; AVX512F-FAST-NEXT: vpermd %zmm14, %zmm11, %zmm0 {%k1}
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm14
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm27
+; AVX512F-FAST-NEXT: vpermd %zmm27, %zmm11, %zmm27
+; AVX512F-FAST-NEXT: vpermd %zmm14, %zmm11, %zmm27 {%k1}
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm14
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm28
+; AVX512F-FAST-NEXT: vpermd %zmm28, %zmm11, %zmm28
+; AVX512F-FAST-NEXT: vpermd %zmm14, %zmm11, %zmm28 {%k1}
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm14
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm29
+; AVX512F-FAST-NEXT: vpermd %zmm29, %zmm11, %zmm29
+; AVX512F-FAST-NEXT: vpermd %zmm14, %zmm11, %zmm29 {%k1}
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm14
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm31
+; AVX512F-FAST-NEXT: vpermd %zmm31, %zmm11, %zmm31
+; AVX512F-FAST-NEXT: vpermd %zmm14, %zmm11, %zmm31 {%k1}
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm14
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm8, %zmm8
+; AVX512F-FAST-NEXT: vpermd %zmm8, %zmm11, %zmm8
+; AVX512F-FAST-NEXT: vpermd %zmm14, %zmm11, %zmm8 {%k1}
+; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm14, %zmm14, %zmm14
; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm2, %zmm2, %zmm18
-; AVX512F-FAST-NEXT: vpermd %zmm18, %zmm22, %zmm18
-; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm11, %zmm18 {%k1}
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm2, %zmm2, %zmm10
-; AVX512F-FAST-NEXT: vpermd %zmm10, %zmm22, %zmm10
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm2, %zmm2, %zmm22
-; AVX512F-FAST-NEXT: vpermd %zmm22, %zmm11, %zmm10 {%k1}
+; AVX512F-FAST-NEXT: vpermd %zmm18, %zmm11, %zmm18
+; AVX512F-FAST-NEXT: vpermd %zmm14, %zmm11, %zmm18 {%k1}
; AVX512F-FAST-NEXT: movb $-86, %al
; AVX512F-FAST-NEXT: kmovw %eax, %k1
; AVX512F-FAST-NEXT: vmovdqa64 %zmm1, %zmm12 {%k1}
; AVX512F-FAST-NEXT: vmovdqu64 (%rsp), %zmm1 # 64-byte Reload
; AVX512F-FAST-NEXT: vmovdqa64 %zmm1, %zmm6 {%k1}
; AVX512F-FAST-NEXT: vmovdqa64 %zmm3, %zmm7 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm16, %zmm14 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm21, %zmm26 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm10, %zmm0 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm16, %zmm27 {%k1}
; AVX512F-FAST-NEXT: vmovdqa64 %zmm24, %zmm13 {%k1}
; AVX512F-FAST-NEXT: vmovdqa64 %zmm25, %zmm15 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm28, %zmm30 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm29, %zmm31 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm20, %zmm28 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm22, %zmm29 {%k1}
; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
; AVX512F-FAST-NEXT: vmovdqa64 %zmm1, %zmm17 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm27, %zmm9 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm23, %zmm5 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm20, %zmm18 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm30, %zmm9 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm26, %zmm31 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm21, %zmm8 {%k1}
; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
; AVX512F-FAST-NEXT: vmovdqa64 %zmm1, %zmm4 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm19, %zmm8 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm0, %zmm10 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm19, %zmm5 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm23, %zmm18 {%k1}
; AVX512F-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm10, (%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm8, 192(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm18, (%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm5, 192(%rax)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm4, 128(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm18, 320(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm5, 256(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm8, 320(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm31, 256(%rax)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm9, 448(%rax)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm17, 384(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm31, 576(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm30, 512(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm29, 576(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm28, 512(%rax)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm15, 704(%rax)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm13, 640(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm26, 832(%rax)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm14, 768(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm27, 832(%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm0, 768(%rax)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm7, 960(%rax)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm6, 896(%rax)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm12, 64(%rax)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll
index 299d74fdf5f177..a781ae5eb89675 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll
@@ -265,16 +265,14 @@ define void @store_i32_stride6_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm6
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm7
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm4, %ymm8
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = [0,4,0,4]
-; AVX2-ONLY-NEXT: # xmm9 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm9, %ymm9
-; AVX2-ONLY-NEXT: vbroadcastf128 {{.*#+}} ymm10 = [0,4,1,5,0,4,1,5]
-; AVX2-ONLY-NEXT: # ymm10 = mem[0,1,0,1]
-; AVX2-ONLY-NEXT: vpermps %ymm6, %ymm10, %ymm10
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1],ymm9[2,3],ymm10[4,5,6,7]
-; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm10 = [0,4,0,4,0,4,0,4]
-; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm10, %ymm10
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm9 = [0,4,0,4,0,4,0,4]
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm9, %ymm10
+; AVX2-ONLY-NEXT: vbroadcastf128 {{.*#+}} ymm11 = [0,4,1,5,0,4,1,5]
+; AVX2-ONLY-NEXT: # ymm11 = mem[0,1,0,1]
+; AVX2-ONLY-NEXT: vpermps %ymm6, %ymm11, %ymm11
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm9, %ymm9
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3],ymm9[4,5],ymm10[6,7]
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm10 = [2,6,2,6,2,6,2,6]
; AVX2-ONLY-NEXT: vpermps %ymm6, %ymm10, %ymm6
; AVX2-ONLY-NEXT: vbroadcastf128 {{.*#+}} ymm10 = [1,5,2,6,1,5,2,6]
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 d24abc59341a5c..e5992acfaa2a77 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
@@ -388,15 +388,13 @@ define void @store_i32_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm10, %ymm10
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm10[0],ymm5[1,2],ymm10[3,4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm9[3,4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm9 = [0,4,0,4]
-; AVX2-SLOW-NEXT: # xmm9 = mem[0,0]
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm9 = [0,4,0,4,0,4,0,4]
; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm9, %ymm6
-; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm9 = [0,4,0,1,0,4,0,1]
-; AVX2-SLOW-NEXT: # ymm9 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm9, %ymm4
+; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm10 = [0,4,0,1,0,4,0,1]
+; AVX2-SLOW-NEXT: # ymm10 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm10, %ymm4
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm6 = [0,4,0,4,0,4,0,4]
-; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm6, %ymm6
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm9, %ymm6
; AVX2-SLOW-NEXT: vbroadcastss (%r10), %ymm7
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6],ymm4[7]
@@ -443,15 +441,13 @@ define void @store_i32_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm4 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm4[0,1,0,1,4,5,4,5]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm4[2,3,4],ymm8[5,6,7]
-; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm5 = [0,4,0,4]
-; AVX2-FAST-NEXT: # xmm5 = mem[0,0]
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm5 = [0,4,0,4,0,4,0,4]
; AVX2-FAST-NEXT: vpermps %ymm3, %ymm5, %ymm3
-; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm5 = [0,4,0,1,0,4,0,1]
-; AVX2-FAST-NEXT: # ymm5 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm5, %ymm2
+; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm8 = [0,4,0,1,0,4,0,1]
+; AVX2-FAST-NEXT: # ymm8 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm8, %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm3 = [0,4,0,4,0,4,0,4]
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpermps %ymm6, %ymm5, %ymm3
; AVX2-FAST-NEXT: vbroadcastss (%r10), %ymm5
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6],ymm2[7]
@@ -499,15 +495,13 @@ define void @store_i32_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm10, %ymm10
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm10[0],ymm5[1,2],ymm10[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm9[3,4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm9 = [0,4,0,4]
-; AVX2-FAST-PERLANE-NEXT: # xmm9 = mem[0,0]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm9 = [0,4,0,4,0,4,0,4]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm9, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm9 = [0,4,0,1,0,4,0,1]
-; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm9, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm10 = [0,4,0,1,0,4,0,1]
+; AVX2-FAST-PERLANE-NEXT: # ymm10 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm10, %ymm4
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3],ymm4[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm6 = [0,4,0,4,0,4,0,4]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm6, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm9, %ymm6
; AVX2-FAST-PERLANE-NEXT: vbroadcastss (%r10), %ymm7
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6],ymm4[7]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll
index 85ffcdf71e3cbf..5c844cf04e1f50 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll
@@ -256,49 +256,41 @@ define void @store_i32_stride8_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vmovaps (%r8), %xmm5
; AVX2-ONLY-NEXT: vmovaps (%r11), %xmm6
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm2
-; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm8
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm7
; AVX2-ONLY-NEXT: vinsertf128 $1, (%r9), %ymm5, %ymm5
-; AVX2-ONLY-NEXT: vinsertf128 $1, (%r10), %ymm6, %ymm7
+; AVX2-ONLY-NEXT: vinsertf128 $1, (%r10), %ymm6, %ymm8
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm6 = [0,4,0,4,0,4,0,4]
-; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm6, %ymm9
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm6, %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = [0,4,0,4]
-; AVX2-ONLY-NEXT: # xmm9 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm9, %ymm9
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm6, %ymm9
+; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm6, %ymm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4,5],ymm9[6,7]
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm10 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm6, %ymm6
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm10[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm9[4,5,6,7]
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm9 = [1,5,1,5,1,5,1,5]
-; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm9, %ymm10
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm9, %ymm9
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm10 = [1,5,1,5]
-; AVX2-ONLY-NEXT: # xmm10 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm10, %ymm10
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm9, %ymm10
+; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm9, %ymm11
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4,5],ymm10[6,7]
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm11 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0,1],xmm11[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm9, %ymm9
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0,1],xmm11[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm10 = [2,6,2,6,2,6,2,6]
-; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm10, %ymm11
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm10, %ymm10
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [2,6,2,6]
-; AVX2-ONLY-NEXT: # xmm11 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm11, %ymm8
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm10, %ymm11
+; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm10, %ymm12
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm8[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm10, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm11[4,5,6,7]
; AVX2-ONLY-NEXT: vbroadcastsd {{.*#+}} ymm4 = [3,7,3,7,3,7,3,7]
-; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm4, %ymm7
-; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm4, %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = [3,7,3,7]
-; AVX2-ONLY-NEXT: # xmm5 = mem[0,0]
-; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm5, %ymm2
+; AVX2-ONLY-NEXT: vpermps %ymm8, %ymm4, %ymm7
+; AVX2-ONLY-NEXT: vpermps %ymm5, %ymm4, %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm7[6,7]
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vpermps %ymm2, %ymm4, %ymm1
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps %ymm0, 96(%rax)
; AVX2-ONLY-NEXT: vmovaps %ymm3, 64(%rax)
; AVX2-ONLY-NEXT: vmovaps %ymm9, 32(%rax)
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 4174fde817472a..9f664f8f165290 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
@@ -4063,181 +4063,184 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512F-FAST-LABEL: store_i8_stride5_vf64:
; AVX512F-FAST: # %bb.0:
; AVX512F-FAST-NEXT: subq $24, %rsp
-; AVX512F-FAST-NEXT: vmovdqa 32(%rsi), %ymm3
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm8 = [128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128,128,128,128]
-; AVX512F-FAST-NEXT: vpshufb %ymm8, %ymm3, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa 32(%rsi), %ymm9
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128,128,128,128]
+; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm9, %ymm0
; AVX512F-FAST-NEXT: vmovdqa 32(%rdi), %ymm1
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = <12,13,128,15,12,13,14,128,12,13,14,15,128,u,u,u,16,128,18,19,16,17,128,19,16,17,18,128,16,17,18,19>
-; AVX512F-FAST-NEXT: vpshufb %ymm11, %ymm1, %ymm2
-; AVX512F-FAST-NEXT: vpor %ymm0, %ymm2, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <12,13,128,15,12,13,14,128,12,13,14,15,128,u,u,u,16,128,18,19,16,17,128,19,16,17,18,128,16,17,18,19>
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm3
+; AVX512F-FAST-NEXT: vpor %ymm0, %ymm3, %ymm0
; AVX512F-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT: vmovdqa 32(%rdi), %xmm0
; AVX512F-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = <8,128,u,7,128,9,128,u,128,u,10,128,12,128,u,11>
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm0, %xmm0
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = <8,128,u,7,128,9,128,u,128,u,10,128,12,128,u,11>
+; AVX512F-FAST-NEXT: vpshufb %xmm4, %xmm0, %xmm0
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm4, %xmm25
; AVX512F-FAST-NEXT: vmovdqa 32(%rsi), %xmm4
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm13 = <128,8,u,128,7,128,9,u,11,u,128,10,128,12,u,128>
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm4, %xmm2
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm4, %xmm19
-; AVX512F-FAST-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = <128,8,u,128,7,128,9,u,11,u,128,10,128,12,u,128>
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm4, %xmm3
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm5, %xmm26
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm4, %xmm20
+; AVX512F-FAST-NEXT: vpor %xmm0, %xmm3, %xmm0
; AVX512F-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa 32(%rcx), %ymm7
+; AVX512F-FAST-NEXT: vmovdqa 32(%rcx), %ymm10
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128,128]
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm7, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa 32(%rdx), %ymm9
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,12,13,128,u,u,u,14,128,u,u,14,15,128,u,u,u,16,128,u,u,16,17,128,u,u,u,18,128,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm15, %ymm9, %ymm4
-; AVX512F-FAST-NEXT: vpor %ymm2, %ymm4, %ymm2
-; AVX512F-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa 32(%rcx), %xmm12
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = <128,6,128,8,u,128,7,128,9,128,11,u,128,10,128,12>
-; AVX512F-FAST-NEXT: vpshufb %xmm4, %xmm12, %xmm2
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm4, %xmm25
-; AVX512F-FAST-NEXT: vmovdqa 32(%rdx), %xmm10
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm6 = <6,128,8,128,u,7,128,9,128,11,128,u,10,128,12,128>
-; AVX512F-FAST-NEXT: vpshufb %xmm6, %xmm10, %xmm4
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm6, %xmm26
-; AVX512F-FAST-NEXT: vpor %xmm2, %xmm4, %xmm2
-; AVX512F-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm3[11,u,u,10,u,12,u,u,u,u,13,u,15,u,u,14,27,u,u,26,u,28,u,u,u,u,29,u,31,u,u,30]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[3,u,5,u,u,4,u,6,u,8,u,u,7,u,9,u,19,u,21,u,u,20,u,22,u,24,u,u,23,u,25,u]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm21
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,zero,ymm1[26],zero,ymm1[28],zero,zero,ymm1[27],zero,ymm1[29],zero,ymm1[31],zero,zero,ymm1[30],zero
+; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm10, %ymm4
+; AVX512F-FAST-NEXT: vmovdqa 32(%rdx), %ymm11
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,12,13,128,u,u,u,14,128,u,u,14,15,128,u,u,u,16,128,u,u,16,17,128,u,u,u,18,128,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm11, %ymm5
+; AVX512F-FAST-NEXT: vpor %ymm4, %ymm5, %ymm4
+; AVX512F-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-FAST-NEXT: vmovdqa 32(%rcx), %xmm14
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = <128,6,128,8,u,128,7,128,9,128,11,u,128,10,128,12>
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm14, %xmm4
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm5, %xmm27
+; AVX512F-FAST-NEXT: vmovdqa 32(%rdx), %xmm13
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm15 = <6,128,8,128,u,7,128,9,128,11,128,u,10,128,12,128>
+; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm13, %xmm5
+; AVX512F-FAST-NEXT: vpor %xmm4, %xmm5, %xmm4
+; AVX512F-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm4 = ymm9[11,u,u,10,u,12,u,u,u,u,13,u,15,u,u,14,27,u,u,26,u,28,u,u,u,u,29,u,31,u,u,30]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm5 = ymm9[3,u,5,u,u,4,u,6,u,8,u,u,7,u,9,u,19,u,21,u,u,20,u,22,u,24,u,u,23,u,25,u]
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm5, %zmm22
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm4 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,zero,ymm1[26],zero,ymm1[28],zero,zero,ymm1[27],zero,ymm1[29],zero,ymm1[31],zero,zero,ymm1[30],zero
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,ymm1[21],zero,zero,ymm1[20],zero,ymm1[22],zero,ymm1[24],zero,zero,ymm1[23],zero,ymm1[25],zero,zero
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm22
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,ymm9[27],zero,zero,ymm9[26],zero,ymm9[28],zero,ymm9[30],zero,zero,ymm9[29],zero,ymm9[31],zero,zero
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,zero,ymm7[19],zero,ymm7[21],zero,zero,ymm7[20],zero,ymm7[22],zero,ymm7[24],zero,zero,ymm7[23],zero
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm23
-; AVX512F-FAST-NEXT: vmovdqa (%rcx), %ymm4
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm4, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa (%rdx), %ymm14
-; AVX512F-FAST-NEXT: vpshufb %ymm15, %ymm14, %ymm1
-; AVX512F-FAST-NEXT: vporq %ymm0, %ymm1, %ymm24
-; AVX512F-FAST-NEXT: vmovdqa (%rsi), %ymm15
-; AVX512F-FAST-NEXT: vpshufb %ymm8, %ymm15, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa (%rdi), %ymm6
-; AVX512F-FAST-NEXT: vpshufb %ymm11, %ymm6, %ymm1
-; AVX512F-FAST-NEXT: vporq %ymm0, %ymm1, %ymm20
-; AVX512F-FAST-NEXT: vmovdqa (%rdi), %xmm0
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm0, %xmm1
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm0, %xmm16
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm1, %zmm23
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,ymm11[27],zero,zero,ymm11[26],zero,ymm11[28],zero,ymm11[30],zero,zero,ymm11[29],zero,ymm11[31],zero,zero
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm4 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,zero,ymm10[19],zero,ymm10[21],zero,zero,ymm10[20],zero,ymm10[22],zero,ymm10[24],zero,zero,ymm10[23],zero
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm4, %zmm24
+; AVX512F-FAST-NEXT: vmovdqa (%rcx), %ymm12
+; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm12, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa (%rdx), %ymm7
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm7, %ymm1
+; AVX512F-FAST-NEXT: vporq %ymm0, %ymm1, %ymm19
+; AVX512F-FAST-NEXT: vmovdqa (%rsi), %ymm5
+; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm5, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa (%rdi), %ymm8
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm8, %ymm1
+; AVX512F-FAST-NEXT: vporq %ymm0, %ymm1, %ymm21
+; AVX512F-FAST-NEXT: vmovdqa (%rdi), %xmm1
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm25, %xmm0
+; AVX512F-FAST-NEXT: vpshufb %xmm0, %xmm1, %xmm0
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm1, %xmm16
; AVX512F-FAST-NEXT: vmovdqa (%rsi), %xmm3
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm3, %xmm2
-; AVX512F-FAST-NEXT: vporq %xmm1, %xmm2, %xmm28
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm26, %xmm1
+; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm3, %xmm2
+; AVX512F-FAST-NEXT: vporq %xmm0, %xmm2, %xmm28
; AVX512F-FAST-NEXT: vmovdqa (%rcx), %xmm1
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm25, %xmm0
-; AVX512F-FAST-NEXT: vpshufb %xmm0, %xmm1, %xmm8
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm27, %xmm0
+; AVX512F-FAST-NEXT: vpshufb %xmm0, %xmm1, %xmm0
; AVX512F-FAST-NEXT: vmovdqa64 %xmm1, %xmm18
; AVX512F-FAST-NEXT: vmovdqa (%rdx), %xmm1
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm26, %xmm0
-; AVX512F-FAST-NEXT: vpshufb %xmm0, %xmm1, %xmm11
+; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm1, %xmm15
; AVX512F-FAST-NEXT: vmovdqa64 %xmm1, %xmm17
-; AVX512F-FAST-NEXT: vporq %xmm8, %xmm11, %xmm29
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = [1,1,2,2,2,2,2,2]
-; AVX512F-FAST-NEXT: vmovdqa 32(%r8), %ymm8
-; AVX512F-FAST-NEXT: vpermd %ymm8, %ymm11, %ymm11
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm25 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
-; AVX512F-FAST-NEXT: vpandnq %ymm11, %ymm25, %ymm11
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [12,128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128]
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm8, %ymm13
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm13, %zmm11, %zmm26
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm31 = <4,u,5,5,5,5,u,6,30,30,30,u,31,31,31,31>
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm30 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0]
-; AVX512F-FAST-NEXT: vmovdqa (%r8), %ymm11
-; AVX512F-FAST-NEXT: vpermd %ymm11, %ymm31, %ymm27
-; AVX512F-FAST-NEXT: vpandnq %ymm27, %ymm30, %ymm27
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm11, %ymm0
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm27, %zmm0, %zmm27
-; AVX512F-FAST-NEXT: vpbroadcastq {{.*#+}} ymm0 = [9,14,11,0,13,10,15,12,9,14,11,0,13,10,15,12,9,14,11,0,13,10,15,12,9,14,11,0,13,10,15,12]
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm7, %ymm7
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm13 = ymm4[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,zero,ymm4[19],zero,ymm4[21],zero,zero,ymm4[20],zero,ymm4[22],zero,ymm4[24],zero,zero,ymm4[23],zero
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm4, %ymm5
-; AVX512F-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [18,19,128,21,128,21,20,128,22,128,24,128,22,23,128,25,18,19,128,21,128,21,20,128,22,128,24,128,22,23,128,25]
-; AVX512F-FAST-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm9, %ymm9
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm14, %ymm4
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm14 = 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
-; AVX512F-FAST-NEXT: vpunpcklbw {{.*#+}} xmm10 = xmm12[0],xmm10[0],xmm12[1],xmm10[1],xmm12[2],xmm10[2],xmm12[3],xmm10[3],xmm12[4],xmm10[4],xmm12[5],xmm10[5],xmm12[6],xmm10[6],xmm12[7],xmm10[7]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm12 = ymm15[3,u,5,u,u,4,u,6,u,8,u,u,7,u,9,u,19,u,21,u,u,20,u,22,u,24,u,u,23,u,25,u]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm15 = ymm15[11,u,u,10,u,12,u,u,u,u,13,u,15,u,u,14,27,u,u,26,u,28,u,u,u,u,29,u,31,u,u,30]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,ymm6[21],zero,zero,ymm6[20],zero,ymm6[22],zero,ymm6[24],zero,zero,ymm6[23],zero,ymm6[25],zero,zero
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,zero,ymm6[26],zero,ymm6[28],zero,zero,ymm6[27],zero,ymm6[29],zero,ymm6[31],zero,zero,ymm6[30],zero
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm19, %xmm1
+; AVX512F-FAST-NEXT: vporq %xmm0, %xmm15, %xmm29
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = [1,1,2,2,2,2,2,2]
+; AVX512F-FAST-NEXT: vmovdqa 32(%r8), %ymm0
+; AVX512F-FAST-NEXT: vpermd %ymm0, %ymm15, %ymm15
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm30 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
+; AVX512F-FAST-NEXT: vpandnq %ymm15, %ymm30, %ymm15
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [12,128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128]
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm9
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm15, %zmm27
+; AVX512F-FAST-NEXT: vmovdqa64 (%r8), %zmm25
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm15 = <4,u,5,5,5,5,u,6,30,30,30,u,31,31,31,31>
+; AVX512F-FAST-NEXT: vpermi2d %zmm25, %zmm0, %zmm15
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = [4,6,5,5,5,5,4,6]
+; AVX512F-FAST-NEXT: vmovdqa (%r8), %ymm1
+; AVX512F-FAST-NEXT: vpermd %ymm1, %ymm9, %ymm9
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm31 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0]
+; AVX512F-FAST-NEXT: vpandnq %ymm9, %ymm31, %ymm9
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm2
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm2, %zmm26
+; AVX512F-FAST-NEXT: vpbroadcastq {{.*#+}} ymm2 = [9,14,11,0,13,10,15,12,9,14,11,0,13,10,15,12,9,14,11,0,13,10,15,12,9,14,11,0,13,10,15,12]
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm10, %ymm9
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm10 = ymm12[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,zero,ymm12[19],zero,ymm12[21],zero,zero,ymm12[20],zero,ymm12[22],zero,ymm12[24],zero,zero,ymm12[23],zero
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm12, %ymm6
+; AVX512F-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm12 = [18,19,128,21,128,21,20,128,22,128,24,128,22,23,128,25,18,19,128,21,128,21,20,128,22,128,24,128,22,23,128,25]
+; AVX512F-FAST-NEXT: # ymm12 = mem[0,1,0,1]
+; AVX512F-FAST-NEXT: vpshufb %ymm12, %ymm11, %ymm11
+; AVX512F-FAST-NEXT: vpshufb %ymm12, %ymm7, %ymm12
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,ymm7[27],zero,zero,ymm7[26],zero,ymm7[28],zero,ymm7[30],zero,zero,ymm7[29],zero,ymm7[31],zero,zero
+; AVX512F-FAST-NEXT: vpunpcklbw {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1],xmm14[2],xmm13[2],xmm14[3],xmm13[3],xmm14[4],xmm13[4],xmm14[5],xmm13[5],xmm14[6],xmm13[6],xmm14[7],xmm13[7]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm14 = ymm5[3,u,5,u,u,4,u,6,u,8,u,u,7,u,9,u,19,u,21,u,u,20,u,22,u,24,u,u,23,u,25,u]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[11,u,u,10,u,12,u,u,u,u,13,u,15,u,u,14,27,u,u,26,u,28,u,u,u,u,29,u,31,u,u,30]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm4 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,ymm8[21],zero,zero,ymm8[20],zero,ymm8[22],zero,ymm8[24],zero,zero,ymm8[23],zero,ymm8[25],zero,zero
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u],zero,zero,ymm8[26],zero,ymm8[28],zero,zero,ymm8[27],zero,ymm8[29],zero,ymm8[31],zero,zero,ymm8[30],zero
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm20, %xmm2
; AVX512F-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX512F-FAST-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm16, %xmm1
-; AVX512F-FAST-NEXT: vpunpcklbw {{.*#+}} xmm3 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = <0,1,4,5,u,2,3,6,7,10,11,u,8,9,12,13>
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm3, %xmm1
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm28, %zmm1, %zmm28
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm1 = ymm7[2,2,3,3]
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm7 = ymm9[2,2,3,3]
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm9 = ymm13[2,2,3,3]
+; AVX512F-FAST-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm16, %xmm2
+; AVX512F-FAST-NEXT: vpunpcklbw {{.*#+}} xmm3 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = <0,1,4,5,u,2,3,6,7,10,11,u,8,9,12,13>
+; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm0
+; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm2
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm28, %zmm2, %zmm28
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm9[2,2,3,3]
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm9 = ymm11[2,2,3,3]
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm10 = ymm10[2,2,3,3]
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm11 = ymm12[2,2,3,3]
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm12 = ymm14[2,2,3,3]
; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,2,3,3]
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm12 = ymm12[2,2,3,3]
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,2,3,3]
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm13 = ymm5[2,2,3,3]
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm14 = ymm14[2,2,3,3]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm18, %xmm3
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm17, %xmm5
-; AVX512F-FAST-NEXT: vpunpcklbw {{.*#+}} xmm5 = 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]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = <2,u,1,0,5,4,u,3,u,7,6,11,10,u,9,8>
-; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm10, %xmm10
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,0,1,1]
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm15 = ymm15[2,2,3,3]
; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm6 = ymm6[2,2,3,3]
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm7 = ymm7[2,2,3,3]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm18, %xmm2
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm17, %xmm14
+; AVX512F-FAST-NEXT: vpunpcklbw {{.*#+}} xmm14 = xmm2[0],xmm14[0],xmm2[1],xmm14[1],xmm2[2],xmm14[2],xmm2[3],xmm14[3],xmm2[4],xmm14[4],xmm2[5],xmm14[5],xmm2[6],xmm14[6],xmm2[7],xmm14[7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = <2,u,1,0,5,4,u,3,u,7,6,11,10,u,9,8>
+; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm13, %xmm13
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm13 = ymm13[0,0,1,1]
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm5 = ymm5[2,2,3,3]
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm8[2,2,3,3]
; AVX512F-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,0,1,1]
-; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm5, %xmm3
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm29, %zmm3, %zmm3
-; AVX512F-FAST-NEXT: vmovdqa64 (%r8), %zmm5
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm5, %zmm11
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} zmm21 = zmm21[2,2,3,3,6,6,7,7]
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} zmm22 = zmm22[2,2,3,3,6,6,7,7]
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm29 = [18374966859431608575,18374966859431608575,18446463693966278400,18446463693966278400,72056498804555775,72056498804555775,18374967950370078975,18374967950370078975]
-; AVX512F-FAST-NEXT: vpternlogq $248, %zmm29, %zmm21, %zmm22
-; AVX512F-FAST-NEXT: vpandq %ymm29, %ymm1, %ymm1
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm7, %zmm1
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} zmm7 = zmm23[2,2,3,3,6,6,7,7]
-; AVX512F-FAST-NEXT: vporq %zmm7, %zmm1, %zmm1
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm7 = [0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255]
-; AVX512F-FAST-NEXT: vpternlogq $226, %zmm22, %zmm7, %zmm1
-; AVX512F-FAST-NEXT: vpermt2d %zmm5, %zmm31, %zmm8
-; AVX512F-FAST-NEXT: vpternlogd $184, %zmm1, %zmm30, %zmm8
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm8, 256(%r9)
-; AVX512F-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: # ymm1 = mem[0,0,1,1]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm1 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: # ymm8 = mem[0,0,1,1]
-; AVX512F-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm8 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm16 = [255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0]
-; AVX512F-FAST-NEXT: vpternlogq $226, %zmm1, %zmm16, %zmm8
-; AVX512F-FAST-NEXT: vpternlogq $248, %zmm25, %zmm8, %zmm26
-; AVX512F-FAST-NEXT: vpor %ymm4, %ymm9, %ymm1
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm24, %zmm1
-; AVX512F-FAST-NEXT: vpternlogq $248, %ymm29, %ymm12, %ymm2
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm20, %zmm2
-; AVX512F-FAST-NEXT: vpternlogq $226, %zmm1, %zmm16, %zmm2
-; AVX512F-FAST-NEXT: vpternlogq $248, %ymm29, %ymm13, %ymm14
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm10, %zmm14, %zmm1
-; AVX512F-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm15, %ymm6
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm6, %zmm0
-; AVX512F-FAST-NEXT: vpternlogq $226, %zmm1, %zmm7, %zmm0
-; AVX512F-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm27
-; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = <6,6,6,u,7,7,7,7,u,8,8,8,8,u,9,9>
-; AVX512F-FAST-NEXT: vpermd %zmm5, %zmm1, %zmm1
-; AVX512F-FAST-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
+; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm14, %xmm2
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm29, %zmm2, %zmm2
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm25, %zmm1
+; AVX512F-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: # ymm14 = mem[0,0,1,1]
+; AVX512F-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm14, %zmm14 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm16 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: # ymm16 = mem[0,0,1,1]
+; AVX512F-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm16, %zmm16 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm17 = [255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0]
+; AVX512F-FAST-NEXT: vpternlogq $226, %zmm14, %zmm17, %zmm16
+; AVX512F-FAST-NEXT: vpternlogq $248, %zmm30, %zmm16, %zmm27
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} zmm14 = zmm22[2,2,3,3,6,6,7,7]
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} zmm16 = zmm23[2,2,3,3,6,6,7,7]
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm18 = [18374966859431608575,18374966859431608575,18446463693966278400,18446463693966278400,72056498804555775,72056498804555775,18374967950370078975,18374967950370078975]
+; AVX512F-FAST-NEXT: vpternlogq $248, %zmm18, %zmm14, %zmm16
+; AVX512F-FAST-NEXT: vpandq %ymm18, %ymm3, %ymm3
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm9, %zmm3
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} zmm9 = zmm24[2,2,3,3,6,6,7,7]
+; AVX512F-FAST-NEXT: vporq %zmm9, %zmm3, %zmm3
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm9 = [0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255]
+; AVX512F-FAST-NEXT: vpternlogq $226, %zmm16, %zmm9, %zmm3
+; AVX512F-FAST-NEXT: vpternlogd $184, %zmm3, %zmm31, %zmm15
+; AVX512F-FAST-NEXT: vpor %ymm10, %ymm11, %ymm3
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm19, %zmm3
+; AVX512F-FAST-NEXT: vpternlogq $248, %ymm18, %ymm12, %ymm4
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm21, %zmm4
+; AVX512F-FAST-NEXT: vpternlogq $226, %zmm3, %zmm17, %zmm4
+; AVX512F-FAST-NEXT: vpternlogq $248, %ymm18, %ymm6, %ymm7
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm13, %zmm7, %zmm3
+; AVX512F-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm5, %ymm8
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm8, %zmm0
+; AVX512F-FAST-NEXT: vpternlogq $226, %zmm3, %zmm9, %zmm0
+; AVX512F-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm26
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = <6,6,6,u,7,7,7,7,u,8,8,8,8,u,9,9>
+; AVX512F-FAST-NEXT: vpermd %zmm25, %zmm3, %zmm3
+; AVX512F-FAST-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm3
; AVX512F-FAST-NEXT: vpermq {{.*#+}} zmm0 = zmm28[0,0,1,1,4,4,5,5]
-; AVX512F-FAST-NEXT: vpermq {{.*#+}} zmm2 = zmm3[0,0,1,1,4,4,5,5]
+; AVX512F-FAST-NEXT: vpermq {{.*#+}} zmm2 = zmm2[0,0,1,1,4,4,5,5]
; AVX512F-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm2
; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = <u,0,0,0,0,u,1,1,9,9,u,10,10,10,10,u>
-; AVX512F-FAST-NEXT: vpermd %zmm11, %zmm0, %zmm0
+; AVX512F-FAST-NEXT: vpermd %zmm1, %zmm0, %zmm0
; AVX512F-FAST-NEXT: vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm0
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm27, 64(%r9)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm26, 64(%r9)
; AVX512F-FAST-NEXT: vmovdqa64 %zmm0, (%r9)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm1, 128(%r9)
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm26, 192(%r9)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm3, 128(%r9)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm15, 256(%r9)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm27, 192(%r9)
; AVX512F-FAST-NEXT: addq $24, %rsp
; AVX512F-FAST-NEXT: vzeroupper
; AVX512F-FAST-NEXT: retq
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 04c1bd53225206..a00cfe5d3d021c 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
@@ -1162,83 +1162,82 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10
-; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovdqa (%rsi), %xmm5
; AVX1-ONLY-NEXT: vmovdqa (%rdx), %xmm6
; AVX1-ONLY-NEXT: vmovdqa (%rcx), %xmm7
; AVX1-ONLY-NEXT: vmovdqa (%r8), %xmm0
; AVX1-ONLY-NEXT: vmovdqa (%r9), %xmm2
; AVX1-ONLY-NEXT: vmovdqa (%r10), %xmm1
-; AVX1-ONLY-NEXT: vpunpcklbw {{.*#+}} xmm9 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3],xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm9[6,7],zero,zero,zero,zero,zero,xmm9[8,9],zero,zero
-; AVX1-ONLY-NEXT: vpunpcklbw {{.*#+}} xmm10 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3],xmm6[4],xmm7[4],xmm6[5],xmm7[5],xmm6[6],xmm7[6],xmm6[7],xmm7[7]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = xmm10[4,5],zero,zero,zero,zero,zero,xmm10[6,7],zero,zero,zero,zero,zero,xmm10[8,9]
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm8, %xmm3
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = zero,zero,xmm10[0,1],zero,zero,zero,zero,zero,xmm10[2,3],zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm9[0,1],zero,zero,zero,zero,zero,xmm9[2,3],zero,zero,zero,zero,zero,xmm9[4,5]
-; AVX1-ONLY-NEXT: vpor %xmm8, %xmm11, %xmm8
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm8, %ymm3
-; AVX1-ONLY-NEXT: vpunpcklbw {{.*#+}} xmm8 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm8[u,u,4,5],zero,xmm8[u,u,u,u,6,7],zero,xmm8[u,u,u,u]
+; AVX1-ONLY-NEXT: vpunpcklbw {{.*#+}} xmm10 = 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]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm4 = zero,zero,zero,zero,zero,xmm10[6,7],zero,zero,zero,zero,zero,xmm10[8,9],zero,zero
+; AVX1-ONLY-NEXT: vpunpcklbw {{.*#+}} xmm11 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3],xmm6[4],xmm7[4],xmm6[5],xmm7[5],xmm6[6],xmm7[6],xmm6[7],xmm7[7]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = xmm11[4,5],zero,zero,zero,zero,zero,xmm11[6,7],zero,zero,zero,zero,zero,xmm11[8,9]
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm8, %xmm4
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = zero,zero,xmm11[0,1],zero,zero,zero,zero,zero,xmm11[2,3],zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm9 = xmm10[0,1],zero,zero,zero,zero,zero,xmm10[2,3],zero,zero,zero,zero,zero,xmm10[4,5]
+; AVX1-ONLY-NEXT: vpor %xmm8, %xmm9, %xmm8
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm8, %ymm4
+; AVX1-ONLY-NEXT: vpunpcklbw {{.*#+}} xmm9 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = xmm9[u,u,4,5],zero,xmm9[u,u,u,u,6,7],zero,xmm9[u,u,u,u]
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm1[u,u],zero,zero,xmm1[2,u,u,u,u],zero,zero,xmm1[3,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm12, %xmm11, %xmm11
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm8[u,u,u,u,0,1,u,u,u,u,u,2,3,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm12, %xmm8, %xmm8
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm9[u,u,u,u,0,1,u,u,u,u,u,2,3,u,u,u]
; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm12 = xmm12[4,5,6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3]
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm12[u,u,u,u,0,1,12,u,u,u,u,7,8,13,u,u]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm12, %ymm11
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm12, %ymm8
; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm12 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
-; AVX1-ONLY-NEXT: vandnps %ymm11, %ymm12, %ymm11
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm11, %ymm3
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm5[u],zero,xmm5[7,u,u,u,u,u],zero,xmm5[8,u,u,u,u,u],zero
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm4[u,7],zero,xmm4[u,u,u,u,u,8],zero,xmm4[u,u,u,u,u,9]
-; AVX1-ONLY-NEXT: vpor %xmm11, %xmm12, %xmm11
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm7[u,u,u],zero,xmm7[7,u,u,u,u,u],zero,xmm7[8,u,u,u,u]
+; AVX1-ONLY-NEXT: vandnps %ymm8, %ymm12, %ymm8
+; AVX1-ONLY-NEXT: vorps %ymm4, %ymm8, %ymm4
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = xmm5[u],zero,xmm5[7,u,u,u,u,u],zero,xmm5[8,u,u,u,u,u],zero
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm3[u,7],zero,xmm3[u,u,u,u,u,8],zero,xmm3[u,u,u,u,u,9]
+; AVX1-ONLY-NEXT: vpor %xmm8, %xmm12, %xmm12
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = xmm7[u,u,u],zero,xmm7[7,u,u,u,u,u],zero,xmm7[8,u,u,u,u]
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm6[u,u,u,7],zero,xmm6[u,u,u,u,u,8],zero,xmm6[u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm8, %xmm13, %xmm13
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} ymm8 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0]
+; AVX1-ONLY-NEXT: vpblendvb %xmm8, %xmm12, %xmm13, %xmm12
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = zero,zero,zero,zero,zero,xmm11[10,11],zero,zero,zero,zero,zero,xmm11[12,13],zero,zero
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm10 = zero,zero,zero,xmm10[10,11],zero,zero,zero,zero,zero,xmm10[12,13],zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpor %xmm11, %xmm10, %xmm10
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm10, %ymm10
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm11 = [0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255]
+; AVX1-ONLY-NEXT: vandps %ymm11, %ymm10, %ymm10
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm2[u,u,u,u,u],zero,xmm2[7,u,u,u,u,u],zero,xmm2[8,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm0[u,u,u,u,u,7],zero,xmm0[u,u,u,u,u,8],zero,xmm0[u,u]
; AVX1-ONLY-NEXT: vpor %xmm12, %xmm13, %xmm12
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm13 = <u,255,255,0,0,u,u,u,255,255,0,0,u,u,u,255>
-; AVX1-ONLY-NEXT: vpblendvb %xmm13, %xmm11, %xmm12, %xmm11
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm10 = zero,zero,zero,zero,zero,xmm10[10,11],zero,zero,zero,zero,zero,xmm10[12,13],zero,zero
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm9 = zero,zero,zero,xmm9[10,11],zero,zero,zero,zero,zero,xmm9[12,13],zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpor %xmm10, %xmm9, %xmm9
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm9, %ymm9
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm10 = [0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255]
-; AVX1-ONLY-NEXT: vandps %ymm10, %ymm9, %ymm9
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm2[u,u,u,u,u],zero,xmm2[7,u,u,u,u,u],zero,xmm2[8,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm0[u,u,u,u,u,7],zero,xmm0[u,u,u,u,u,8],zero,xmm0[u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = zero,xmm12[u,u,u,u,5,6],zero,xmm12[u,u,u,u,12,13],zero,xmm12[u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm1[6,u,u,u,u],zero,zero,xmm1[7,u,u,u,u],zero,zero,xmm1[8,u]
+; AVX1-ONLY-NEXT: vpor %xmm13, %xmm12, %xmm12
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm9 = xmm9[8,9],zero,xmm9[u,u,u,u,10,11],zero,xmm9[u,u,u,u,12,13]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = zero,zero,xmm1[4,u,u,u,u],zero,zero,xmm1[5,u,u,u,u],zero,zero
+; AVX1-ONLY-NEXT: vpor %xmm13, %xmm9, %xmm9
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm9, %ymm9
+; AVX1-ONLY-NEXT: vandnps %ymm9, %ymm11, %ymm9
+; AVX1-ONLY-NEXT: vorps %ymm9, %ymm10, %ymm9
+; AVX1-ONLY-NEXT: vpunpckhbw {{.*#+}} xmm10 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = zero,zero,zero,zero,xmm10[8,9],zero,zero,zero,zero,zero,xmm10[10,11],zero,zero,zero
+; AVX1-ONLY-NEXT: vpunpckhbw {{.*#+}} xmm12 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm12[6],zero,zero,zero,zero,zero,xmm12[9,8],zero,zero,zero,zero,zero,xmm12[11,10],zero
; AVX1-ONLY-NEXT: vpor %xmm11, %xmm12, %xmm11
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = zero,xmm11[u,u,u,u,5,6],zero,xmm11[u,u,u,u,12,13],zero,xmm11[u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm1[6,u,u,u,u],zero,zero,xmm1[7,u,u,u,u],zero,zero,xmm1[8,u]
-; AVX1-ONLY-NEXT: vpor %xmm12, %xmm11, %xmm11
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[8,9],zero,xmm8[u,u,u,u,10,11],zero,xmm8[u,u,u,u,12,13]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = zero,zero,xmm1[4,u,u,u,u],zero,zero,xmm1[5,u,u,u,u],zero,zero
-; AVX1-ONLY-NEXT: vpor %xmm12, %xmm8, %xmm8
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm8, %ymm8
-; AVX1-ONLY-NEXT: vandnps %ymm8, %ymm10, %ymm8
-; AVX1-ONLY-NEXT: vorps %ymm8, %ymm9, %ymm8
-; AVX1-ONLY-NEXT: vpunpckhbw {{.*#+}} xmm9 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm10 = zero,zero,zero,zero,xmm9[8,9],zero,zero,zero,zero,zero,xmm9[10,11],zero,zero,zero
-; AVX1-ONLY-NEXT: vpunpckhbw {{.*#+}} xmm11 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[6],zero,zero,zero,zero,zero,xmm11[9,8],zero,zero,zero,zero,zero,xmm11[11,10],zero
-; AVX1-ONLY-NEXT: vpor %xmm10, %xmm11, %xmm10
; AVX1-ONLY-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm6[8],xmm7[8],xmm6[9],xmm7[9],xmm6[10],xmm7[10],xmm6[11],xmm7[11],xmm6[12],xmm7[12],xmm6[13],xmm7[13],xmm6[14],xmm7[14],xmm6[15],xmm7[15]
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = zero,xmm6[2,3],zero,zero,zero,zero,zero,xmm6[4,5],zero,zero,zero,zero,zero,xmm6[6]
-; AVX1-ONLY-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[2],zero,zero,zero,zero,zero,xmm4[5,4],zero,zero,zero,zero,zero,xmm4[7,6],zero
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm4, %xmm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm5[8],xmm3[8],xmm5[9],xmm3[9],xmm5[10],xmm3[10],xmm5[11],xmm3[11],xmm5[12],xmm3[12],xmm5[13],xmm3[13],xmm5[14],xmm3[14],xmm5[15],xmm3[15]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[2],zero,zero,zero,zero,zero,xmm3[5,4],zero,zero,zero,zero,zero,xmm3[7,6],zero
+; AVX1-ONLY-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm3, %ymm3
; AVX1-ONLY-NEXT: vpunpckhbw {{.*#+}} xmm5 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = xmm5[u,6,7],zero,xmm5[u,u,u,u,8,9],zero,xmm5[u,u,u,u,10]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm10 = xmm1[u],zero,zero,xmm1[11,u,u,u,u],zero,zero,xmm1[12,u,u,u,u],zero
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm10, %xmm7
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm1[u],zero,zero,xmm1[11,u,u,u,u],zero,zero,xmm1[12,u,u,u,u],zero
+; AVX1-ONLY-NEXT: vpor %xmm7, %xmm11, %xmm7
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,2,3],zero,xmm5[u,u,u,u,4,5],zero,xmm5[u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm10 = xmm1[u,u,u],zero,zero,xmm1[9,u,u,u,u],zero,zero,xmm1[10,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm5, %xmm10, %xmm5
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm1[u,u,u],zero,zero,xmm1[9,u,u,u,u],zero,zero,xmm1[10,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm5, %xmm11, %xmm5
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm5, %ymm5
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm7 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0]
-; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm7, %ymm5
-; AVX1-ONLY-NEXT: vorps %ymm5, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm8, %ymm5
+; AVX1-ONLY-NEXT: vorps %ymm5, %ymm3, %ymm3
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,zero,zero,xmm6[12,13],zero,zero,zero,zero,zero,xmm6[14,15],zero,zero,zero
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,xmm9[12,13],zero,zero,zero,zero,zero,xmm9[14,15],zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm6 = zero,zero,xmm10[12,13],zero,zero,zero,zero,zero,xmm10[14,15],zero,zero,zero,zero,zero
; AVX1-ONLY-NEXT: vpor %xmm5, %xmm6, %xmm5
; AVX1-ONLY-NEXT: vpunpckhbw {{.*#+}} xmm0 = xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[10],zero,xmm0[u,u,u,u,13,12],zero,xmm0[u,u,u,u,15,14],zero
@@ -1247,9 +1246,9 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm1 = [0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0]
; AVX1-ONLY-NEXT: vpblendvb %xmm1, %xmm5, %xmm0, %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, 96(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm9, 32(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rax)
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
More information about the llvm-commits
mailing list