[llvm] 4f41ea2 - [X86] lowerShuffleAsVTRUNC - bit shift the offset elements into place instead of shuffle
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Wed Dec 14 03:41:43 PST 2022
Author: Simon Pilgrim
Date: 2022-12-14T11:41:14Z
New Revision: 4f41ea20167c8ae5e12887b95104a4697a430203
URL: https://github.com/llvm/llvm-project/commit/4f41ea20167c8ae5e12887b95104a4697a430203
DIFF: https://github.com/llvm/llvm-project/commit/4f41ea20167c8ae5e12887b95104a4697a430203.diff
LOG: [X86] lowerShuffleAsVTRUNC - bit shift the offset elements into place instead of shuffle
This helps avoid issues on non-BWI targets which can end up splitting the shuffles to 2 x 256-bit bitshifts of a smaller scalar width
Added:
Modified:
llvm/lib/Target/X86/X86ISelLowering.cpp
llvm/test/CodeGen/X86/shuffle-strided-with-offset-256.ll
llvm/test/CodeGen/X86/shuffle-strided-with-offset-512.ll
llvm/test/CodeGen/X86/shuffle-vs-trunc-512.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll
llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll
llvm/test/CodeGen/X86/x86-interleaved-access.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index c003daea392b..c627d366812b 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -12579,18 +12579,17 @@ static SDValue lowerShuffleAsVTRUNC(const SDLoc &DL, MVT VT, SDValue V1,
MVT ConcatVT = MVT::getVectorVT(VT.getScalarType(), NumElts * 2);
SDValue Src = DAG.getNode(ISD::CONCAT_VECTORS, DL, ConcatVT, V1, V2);
- // Move the offset'd elements into place for the truncation.
- if (Offset) {
- SmallVector<int, 32> OffsetMask(NumElts * 2, -1);
- for (unsigned I = 0, E = NumElts * 2; I < E; I += Scale)
- OffsetMask[I] = I + Offset;
- Src = DAG.getVectorShuffle(ConcatVT, DL, Src, DAG.getUNDEF(ConcatVT),
- OffsetMask);
- }
-
MVT SrcSVT = MVT::getIntegerVT(SrcEltBits);
MVT SrcVT = MVT::getVectorVT(SrcSVT, NumSrcElts);
Src = DAG.getBitcast(SrcVT, Src);
+
+ // Shift the offset'd elements into place for the truncation.
+ // TODO: Use getTargetVShiftByConstNode.
+ if (Offset)
+ Src = DAG.getNode(
+ X86ISD::VSRLI, DL, SrcVT, Src,
+ DAG.getTargetConstant(Offset * EltSizeInBits, DL, MVT::i8));
+
return getAVX512TruncNode(DL, VT, Src, Subtarget, DAG, !UndefUppers);
}
}
diff --git a/llvm/test/CodeGen/X86/shuffle-strided-with-offset-256.ll b/llvm/test/CodeGen/X86/shuffle-strided-with-offset-256.ll
index 02c66e35776e..52dd300f4efe 100644
--- a/llvm/test/CodeGen/X86/shuffle-strided-with-offset-256.ll
+++ b/llvm/test/CodeGen/X86/shuffle-strided-with-offset-256.ll
@@ -149,7 +149,7 @@ define void @shuffle_v32i8_to_v8i8_1(ptr %L, ptr %S) nounwind {
; AVX512F-LABEL: shuffle_v32i8_to_v8i8_1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT: vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT: vpsrld $8, %ymm0, %ymm0
; AVX512F-NEXT: vpmovdb %zmm0, %xmm0
; AVX512F-NEXT: vmovq %xmm0, (%rsi)
; AVX512F-NEXT: vzeroupper
@@ -157,8 +157,7 @@ define void @shuffle_v32i8_to_v8i8_1(ptr %L, ptr %S) nounwind {
;
; AVX512VL-LABEL: shuffle_v32i8_to_v8i8_1:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512VL-NEXT: vpsrlw $8, %ymm0, %ymm0
+; AVX512VL-NEXT: vpsrld $8, (%rdi), %ymm0
; AVX512VL-NEXT: vpmovdb %ymm0, (%rsi)
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
@@ -166,7 +165,7 @@ define void @shuffle_v32i8_to_v8i8_1(ptr %L, ptr %S) nounwind {
; AVX512BW-LABEL: shuffle_v32i8_to_v8i8_1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512BW-NEXT: vpsrlw $8, %ymm0, %ymm0
+; AVX512BW-NEXT: vpsrld $8, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovdb %zmm0, %xmm0
; AVX512BW-NEXT: vmovq %xmm0, (%rsi)
; AVX512BW-NEXT: vzeroupper
@@ -174,7 +173,7 @@ define void @shuffle_v32i8_to_v8i8_1(ptr %L, ptr %S) nounwind {
;
; AVX512BWVL-LABEL: shuffle_v32i8_to_v8i8_1:
; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrlw $8, (%rdi), %ymm0
+; AVX512BWVL-NEXT: vpsrld $8, (%rdi), %ymm0
; AVX512BWVL-NEXT: vpmovdb %ymm0, (%rsi)
; AVX512BWVL-NEXT: vzeroupper
; AVX512BWVL-NEXT: retq
@@ -317,7 +316,7 @@ define void @shuffle_v16i16_to_v4i16_1(ptr %L, ptr %S) nounwind {
; AVX512F-LABEL: shuffle_v16i16_to_v4i16_1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT: vpsrld $16, %ymm0, %ymm0
+; AVX512F-NEXT: vpsrlq $16, %ymm0, %ymm0
; AVX512F-NEXT: vpmovqw %zmm0, %xmm0
; AVX512F-NEXT: vmovq %xmm0, (%rsi)
; AVX512F-NEXT: vzeroupper
@@ -325,7 +324,7 @@ define void @shuffle_v16i16_to_v4i16_1(ptr %L, ptr %S) nounwind {
;
; AVX512VL-LABEL: shuffle_v16i16_to_v4i16_1:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpsrld $16, (%rdi), %ymm0
+; AVX512VL-NEXT: vpsrlq $16, (%rdi), %ymm0
; AVX512VL-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
@@ -333,7 +332,7 @@ define void @shuffle_v16i16_to_v4i16_1(ptr %L, ptr %S) nounwind {
; AVX512BW-LABEL: shuffle_v16i16_to_v4i16_1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512BW-NEXT: vpsrld $16, %ymm0, %ymm0
+; AVX512BW-NEXT: vpsrlq $16, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovqw %zmm0, %xmm0
; AVX512BW-NEXT: vmovq %xmm0, (%rsi)
; AVX512BW-NEXT: vzeroupper
@@ -341,7 +340,7 @@ define void @shuffle_v16i16_to_v4i16_1(ptr %L, ptr %S) nounwind {
;
; AVX512BWVL-LABEL: shuffle_v16i16_to_v4i16_1:
; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrld $16, (%rdi), %ymm0
+; AVX512BWVL-NEXT: vpsrlq $16, (%rdi), %ymm0
; AVX512BWVL-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512BWVL-NEXT: vzeroupper
; AVX512BWVL-NEXT: retq
@@ -385,7 +384,8 @@ define void @shuffle_v16i16_to_v4i16_2(ptr %L, ptr %S) nounwind {
;
; AVX512F-LABEL: shuffle_v16i16_to_v4i16_2:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpshufd {{.*#+}} ymm0 = mem[1,1,3,3,5,5,7,7]
+; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512F-NEXT: vpsrlq $32, %ymm0, %ymm0
; AVX512F-NEXT: vpmovqw %zmm0, %xmm0
; AVX512F-NEXT: vmovq %xmm0, (%rsi)
; AVX512F-NEXT: vzeroupper
@@ -393,14 +393,15 @@ define void @shuffle_v16i16_to_v4i16_2(ptr %L, ptr %S) nounwind {
;
; AVX512VL-LABEL: shuffle_v16i16_to_v4i16_2:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpshufd {{.*#+}} ymm0 = mem[1,1,3,3,5,5,7,7]
+; AVX512VL-NEXT: vpsrlq $32, (%rdi), %ymm0
; AVX512VL-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: shuffle_v16i16_to_v4i16_2:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpshufd {{.*#+}} ymm0 = mem[1,1,3,3,5,5,7,7]
+; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512BW-NEXT: vpsrlq $32, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovqw %zmm0, %xmm0
; AVX512BW-NEXT: vmovq %xmm0, (%rsi)
; AVX512BW-NEXT: vzeroupper
@@ -408,7 +409,7 @@ define void @shuffle_v16i16_to_v4i16_2(ptr %L, ptr %S) nounwind {
;
; AVX512BWVL-LABEL: shuffle_v16i16_to_v4i16_2:
; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpshufd {{.*#+}} ymm0 = mem[1,1,3,3,5,5,7,7]
+; AVX512BWVL-NEXT: vpsrlq $32, (%rdi), %ymm0
; AVX512BWVL-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512BWVL-NEXT: vzeroupper
; AVX512BWVL-NEXT: retq
@@ -502,7 +503,7 @@ define void @shuffle_v32i8_to_v4i8_1(ptr %L, ptr %S) nounwind {
; AVX512F-LABEL: shuffle_v32i8_to_v4i8_1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT: vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT: vpsrlq $8, %ymm0, %ymm0
; AVX512F-NEXT: vpmovqb %zmm0, %xmm0
; AVX512F-NEXT: vmovd %xmm0, (%rsi)
; AVX512F-NEXT: vzeroupper
@@ -510,8 +511,7 @@ define void @shuffle_v32i8_to_v4i8_1(ptr %L, ptr %S) nounwind {
;
; AVX512VL-LABEL: shuffle_v32i8_to_v4i8_1:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512VL-NEXT: vpsrlw $8, %ymm0, %ymm0
+; AVX512VL-NEXT: vpsrlq $8, (%rdi), %ymm0
; AVX512VL-NEXT: vpmovqb %ymm0, (%rsi)
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
@@ -519,7 +519,7 @@ define void @shuffle_v32i8_to_v4i8_1(ptr %L, ptr %S) nounwind {
; AVX512BW-LABEL: shuffle_v32i8_to_v4i8_1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512BW-NEXT: vpsrlw $8, %ymm0, %ymm0
+; AVX512BW-NEXT: vpsrlq $8, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovqb %zmm0, %xmm0
; AVX512BW-NEXT: vmovd %xmm0, (%rsi)
; AVX512BW-NEXT: vzeroupper
@@ -527,7 +527,7 @@ define void @shuffle_v32i8_to_v4i8_1(ptr %L, ptr %S) nounwind {
;
; AVX512BWVL-LABEL: shuffle_v32i8_to_v4i8_1:
; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrlw $8, (%rdi), %ymm0
+; AVX512BWVL-NEXT: vpsrlq $8, (%rdi), %ymm0
; AVX512BWVL-NEXT: vpmovqb %ymm0, (%rsi)
; AVX512BWVL-NEXT: vzeroupper
; AVX512BWVL-NEXT: retq
@@ -552,7 +552,7 @@ define void @shuffle_v32i8_to_v4i8_2(ptr %L, ptr %S) nounwind {
; AVX512F-LABEL: shuffle_v32i8_to_v4i8_2:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT: vpsrld $16, %ymm0, %ymm0
+; AVX512F-NEXT: vpsrlq $16, %ymm0, %ymm0
; AVX512F-NEXT: vpmovqb %zmm0, %xmm0
; AVX512F-NEXT: vmovd %xmm0, (%rsi)
; AVX512F-NEXT: vzeroupper
@@ -560,7 +560,7 @@ define void @shuffle_v32i8_to_v4i8_2(ptr %L, ptr %S) nounwind {
;
; AVX512VL-LABEL: shuffle_v32i8_to_v4i8_2:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpsrld $16, (%rdi), %ymm0
+; AVX512VL-NEXT: vpsrlq $16, (%rdi), %ymm0
; AVX512VL-NEXT: vpmovqb %ymm0, (%rsi)
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
@@ -568,7 +568,7 @@ define void @shuffle_v32i8_to_v4i8_2(ptr %L, ptr %S) nounwind {
; AVX512BW-LABEL: shuffle_v32i8_to_v4i8_2:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512BW-NEXT: vpsrld $16, %ymm0, %ymm0
+; AVX512BW-NEXT: vpsrlq $16, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovqb %zmm0, %xmm0
; AVX512BW-NEXT: vmovd %xmm0, (%rsi)
; AVX512BW-NEXT: vzeroupper
@@ -576,7 +576,7 @@ define void @shuffle_v32i8_to_v4i8_2(ptr %L, ptr %S) nounwind {
;
; AVX512BWVL-LABEL: shuffle_v32i8_to_v4i8_2:
; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrld $16, (%rdi), %ymm0
+; AVX512BWVL-NEXT: vpsrlq $16, (%rdi), %ymm0
; AVX512BWVL-NEXT: vpmovqb %ymm0, (%rsi)
; AVX512BWVL-NEXT: vzeroupper
; AVX512BWVL-NEXT: retq
@@ -601,7 +601,7 @@ define void @shuffle_v32i8_to_v4i8_3(ptr %L, ptr %S) nounwind {
; AVX512F-LABEL: shuffle_v32i8_to_v4i8_3:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT: vpsrld $24, %ymm0, %ymm0
+; AVX512F-NEXT: vpsrlq $24, %ymm0, %ymm0
; AVX512F-NEXT: vpmovqb %zmm0, %xmm0
; AVX512F-NEXT: vmovd %xmm0, (%rsi)
; AVX512F-NEXT: vzeroupper
@@ -609,7 +609,7 @@ define void @shuffle_v32i8_to_v4i8_3(ptr %L, ptr %S) nounwind {
;
; AVX512VL-LABEL: shuffle_v32i8_to_v4i8_3:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpsrld $24, (%rdi), %ymm0
+; AVX512VL-NEXT: vpsrlq $24, (%rdi), %ymm0
; AVX512VL-NEXT: vpmovqb %ymm0, (%rsi)
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
@@ -617,7 +617,7 @@ define void @shuffle_v32i8_to_v4i8_3(ptr %L, ptr %S) nounwind {
; AVX512BW-LABEL: shuffle_v32i8_to_v4i8_3:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512BW-NEXT: vpsrld $24, %ymm0, %ymm0
+; AVX512BW-NEXT: vpsrlq $24, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovqb %zmm0, %xmm0
; AVX512BW-NEXT: vmovd %xmm0, (%rsi)
; AVX512BW-NEXT: vzeroupper
@@ -625,7 +625,7 @@ define void @shuffle_v32i8_to_v4i8_3(ptr %L, ptr %S) nounwind {
;
; AVX512BWVL-LABEL: shuffle_v32i8_to_v4i8_3:
; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrld $24, (%rdi), %ymm0
+; AVX512BWVL-NEXT: vpsrlq $24, (%rdi), %ymm0
; AVX512BWVL-NEXT: vpmovqb %ymm0, (%rsi)
; AVX512BWVL-NEXT: vzeroupper
; AVX512BWVL-NEXT: retq
@@ -649,7 +649,8 @@ define void @shuffle_v32i8_to_v4i8_4(ptr %L, ptr %S) nounwind {
;
; AVX512F-LABEL: shuffle_v32i8_to_v4i8_4:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpshufd {{.*#+}} ymm0 = mem[1,1,3,3,5,5,7,7]
+; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512F-NEXT: vpsrlq $32, %ymm0, %ymm0
; AVX512F-NEXT: vpmovqb %zmm0, %xmm0
; AVX512F-NEXT: vmovd %xmm0, (%rsi)
; AVX512F-NEXT: vzeroupper
@@ -657,14 +658,15 @@ define void @shuffle_v32i8_to_v4i8_4(ptr %L, ptr %S) nounwind {
;
; AVX512VL-LABEL: shuffle_v32i8_to_v4i8_4:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpshufd {{.*#+}} ymm0 = mem[1,1,3,3,5,5,7,7]
+; AVX512VL-NEXT: vpsrlq $32, (%rdi), %ymm0
; AVX512VL-NEXT: vpmovqb %ymm0, (%rsi)
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: shuffle_v32i8_to_v4i8_4:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpshufd {{.*#+}} ymm0 = mem[1,1,3,3,5,5,7,7]
+; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512BW-NEXT: vpsrlq $32, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovqb %zmm0, %xmm0
; AVX512BW-NEXT: vmovd %xmm0, (%rsi)
; AVX512BW-NEXT: vzeroupper
@@ -672,7 +674,7 @@ define void @shuffle_v32i8_to_v4i8_4(ptr %L, ptr %S) nounwind {
;
; AVX512BWVL-LABEL: shuffle_v32i8_to_v4i8_4:
; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpshufd {{.*#+}} ymm0 = mem[1,1,3,3,5,5,7,7]
+; AVX512BWVL-NEXT: vpsrlq $32, (%rdi), %ymm0
; AVX512BWVL-NEXT: vpmovqb %ymm0, (%rsi)
; AVX512BWVL-NEXT: vzeroupper
; AVX512BWVL-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/shuffle-strided-with-offset-512.ll b/llvm/test/CodeGen/X86/shuffle-strided-with-offset-512.ll
index 6938b7700eb7..489ee1c9c5be 100644
--- a/llvm/test/CodeGen/X86/shuffle-strided-with-offset-512.ll
+++ b/llvm/test/CodeGen/X86/shuffle-strided-with-offset-512.ll
@@ -86,41 +86,12 @@ define void @shuffle_v16i32_to_v8i32_1(ptr %L, ptr %S) nounwind {
}
define void @shuffle_v64i8_to_v16i8_1(ptr %L, ptr %S) nounwind {
-; AVX512F-LABEL: shuffle_v64i8_to_v16i8_1:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT: vmovdqa 32(%rdi), %ymm1
-; AVX512F-NEXT: vpsrlw $8, %ymm0, %ymm0
-; AVX512F-NEXT: vpsrlw $8, %ymm1, %ymm1
-; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512F-NEXT: vpmovdb %zmm0, (%rsi)
-; AVX512F-NEXT: vzeroupper
-; AVX512F-NEXT: retq
-;
-; AVX512VL-LABEL: shuffle_v64i8_to_v16i8_1:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512VL-NEXT: vmovdqa 32(%rdi), %ymm1
-; AVX512VL-NEXT: vpsrlw $8, %ymm0, %ymm0
-; AVX512VL-NEXT: vpsrlw $8, %ymm1, %ymm1
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512VL-NEXT: vpmovdb %zmm0, (%rsi)
-; AVX512VL-NEXT: vzeroupper
-; AVX512VL-NEXT: retq
-;
-; AVX512BW-LABEL: shuffle_v64i8_to_v16i8_1:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpsrlw $8, (%rdi), %zmm0
-; AVX512BW-NEXT: vpmovdb %zmm0, (%rsi)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
-;
-; AVX512BWVL-LABEL: shuffle_v64i8_to_v16i8_1:
-; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrlw $8, (%rdi), %zmm0
-; AVX512BWVL-NEXT: vpmovdb %zmm0, (%rsi)
-; AVX512BWVL-NEXT: vzeroupper
-; AVX512BWVL-NEXT: retq
+; AVX512-LABEL: shuffle_v64i8_to_v16i8_1:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpsrld $8, (%rdi), %zmm0
+; AVX512-NEXT: vpmovdb %zmm0, (%rsi)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
%vec = load <64 x i8>, ptr %L
%strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
store <16 x i8> %strided.vec, ptr %S
@@ -128,37 +99,12 @@ define void @shuffle_v64i8_to_v16i8_1(ptr %L, ptr %S) nounwind {
}
define void @shuffle_v64i8_to_v16i8_2(ptr %L, ptr %S) nounwind {
-; AVX512F-LABEL: shuffle_v64i8_to_v16i8_2:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, 32(%rdi), %zmm0, %zmm0
-; AVX512F-NEXT: vpsrld $16, %zmm0, %zmm0
-; AVX512F-NEXT: vpmovdb %zmm0, (%rsi)
-; AVX512F-NEXT: vzeroupper
-; AVX512F-NEXT: retq
-;
-; AVX512VL-LABEL: shuffle_v64i8_to_v16i8_2:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512VL-NEXT: vinserti64x4 $1, 32(%rdi), %zmm0, %zmm0
-; AVX512VL-NEXT: vpsrld $16, %zmm0, %zmm0
-; AVX512VL-NEXT: vpmovdb %zmm0, (%rsi)
-; AVX512VL-NEXT: vzeroupper
-; AVX512VL-NEXT: retq
-;
-; AVX512BW-LABEL: shuffle_v64i8_to_v16i8_2:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpsrld $16, (%rdi), %zmm0
-; AVX512BW-NEXT: vpmovdb %zmm0, (%rsi)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
-;
-; AVX512BWVL-LABEL: shuffle_v64i8_to_v16i8_2:
-; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrld $16, (%rdi), %zmm0
-; AVX512BWVL-NEXT: vpmovdb %zmm0, (%rsi)
-; AVX512BWVL-NEXT: vzeroupper
-; AVX512BWVL-NEXT: retq
+; AVX512-LABEL: shuffle_v64i8_to_v16i8_2:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpsrld $16, (%rdi), %zmm0
+; AVX512-NEXT: vpmovdb %zmm0, (%rsi)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
%vec = load <64 x i8>, ptr %L
%strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
store <16 x i8> %strided.vec, ptr %S
@@ -181,7 +127,7 @@ define void @shuffle_v64i8_to_v16i8_3(ptr %L, ptr %S) nounwind {
define void @shuffle_v32i16_to_v8i16_1(ptr %L, ptr %S) nounwind {
; AVX512-LABEL: shuffle_v32i16_to_v8i16_1:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpsrld $16, (%rdi), %zmm0
+; AVX512-NEXT: vpsrlq $16, (%rdi), %zmm0
; AVX512-NEXT: vpmovqw %zmm0, (%rsi)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
@@ -194,7 +140,7 @@ define void @shuffle_v32i16_to_v8i16_1(ptr %L, ptr %S) nounwind {
define void @shuffle_v32i16_to_v8i16_2(ptr %L, ptr %S) nounwind {
; AVX512-LABEL: shuffle_v32i16_to_v8i16_2:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpshufd {{.*#+}} zmm0 = mem[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512-NEXT: vpsrlq $32, (%rdi), %zmm0
; AVX512-NEXT: vpmovqw %zmm0, (%rsi)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
@@ -218,43 +164,12 @@ define void @shuffle_v32i16_to_v8i16_3(ptr %L, ptr %S) nounwind {
}
define void @shuffle_v64i8_to_v8i8_1(ptr %L, ptr %S) nounwind {
-; AVX512F-LABEL: shuffle_v64i8_to_v8i8_1:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT: vmovdqa 32(%rdi), %ymm1
-; AVX512F-NEXT: vpsrlw $8, %ymm0, %ymm0
-; AVX512F-NEXT: vpsrlw $8, %ymm1, %ymm1
-; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512F-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512F-NEXT: vmovq %xmm0, (%rsi)
-; AVX512F-NEXT: vzeroupper
-; AVX512F-NEXT: retq
-;
-; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_1:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512VL-NEXT: vmovdqa 32(%rdi), %ymm1
-; AVX512VL-NEXT: vpsrlw $8, %ymm0, %ymm0
-; AVX512VL-NEXT: vpsrlw $8, %ymm1, %ymm1
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512VL-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512VL-NEXT: vmovq %xmm0, (%rsi)
-; AVX512VL-NEXT: vzeroupper
-; AVX512VL-NEXT: retq
-;
-; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_1:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpsrlw $8, (%rdi), %zmm0
-; AVX512BW-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
-;
-; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_1:
-; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrlw $8, (%rdi), %zmm0
-; AVX512BWVL-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BWVL-NEXT: vzeroupper
-; AVX512BWVL-NEXT: retq
+; AVX512-LABEL: shuffle_v64i8_to_v8i8_1:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpsrlq $8, (%rdi), %zmm0
+; AVX512-NEXT: vpmovqb %zmm0, (%rsi)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
%vec = load <64 x i8>, ptr %L
%strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57>
store <8 x i8> %strided.vec, ptr %S
@@ -262,39 +177,12 @@ define void @shuffle_v64i8_to_v8i8_1(ptr %L, ptr %S) nounwind {
}
define void @shuffle_v64i8_to_v8i8_2(ptr %L, ptr %S) nounwind {
-; AVX512F-LABEL: shuffle_v64i8_to_v8i8_2:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, 32(%rdi), %zmm0, %zmm0
-; AVX512F-NEXT: vpsrld $16, %zmm0, %zmm0
-; AVX512F-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512F-NEXT: vmovq %xmm0, (%rsi)
-; AVX512F-NEXT: vzeroupper
-; AVX512F-NEXT: retq
-;
-; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_2:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512VL-NEXT: vinserti64x4 $1, 32(%rdi), %zmm0, %zmm0
-; AVX512VL-NEXT: vpsrld $16, %zmm0, %zmm0
-; AVX512VL-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512VL-NEXT: vmovq %xmm0, (%rsi)
-; AVX512VL-NEXT: vzeroupper
-; AVX512VL-NEXT: retq
-;
-; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_2:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpsrld $16, (%rdi), %zmm0
-; AVX512BW-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
-;
-; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_2:
-; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrld $16, (%rdi), %zmm0
-; AVX512BWVL-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BWVL-NEXT: vzeroupper
-; AVX512BWVL-NEXT: retq
+; AVX512-LABEL: shuffle_v64i8_to_v8i8_2:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpsrlq $16, (%rdi), %zmm0
+; AVX512-NEXT: vpmovqb %zmm0, (%rsi)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
%vec = load <64 x i8>, ptr %L
%strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58>
store <8 x i8> %strided.vec, ptr %S
@@ -302,35 +190,12 @@ define void @shuffle_v64i8_to_v8i8_2(ptr %L, ptr %S) nounwind {
}
define void @shuffle_v64i8_to_v8i8_3(ptr %L, ptr %S) nounwind {
-; AVX512F-LABEL: shuffle_v64i8_to_v8i8_3:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpsrld $24, (%rdi), %zmm0
-; AVX512F-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512F-NEXT: vmovq %xmm0, (%rsi)
-; AVX512F-NEXT: vzeroupper
-; AVX512F-NEXT: retq
-;
-; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_3:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpsrld $24, (%rdi), %zmm0
-; AVX512VL-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512VL-NEXT: vmovq %xmm0, (%rsi)
-; AVX512VL-NEXT: vzeroupper
-; AVX512VL-NEXT: retq
-;
-; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_3:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpsrld $24, (%rdi), %zmm0
-; AVX512BW-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
-;
-; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_3:
-; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrld $24, (%rdi), %zmm0
-; AVX512BWVL-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BWVL-NEXT: vzeroupper
-; AVX512BWVL-NEXT: retq
+; AVX512-LABEL: shuffle_v64i8_to_v8i8_3:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpsrlq $24, (%rdi), %zmm0
+; AVX512-NEXT: vpmovqb %zmm0, (%rsi)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
%vec = load <64 x i8>, ptr %L
%strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 51, i32 59>
store <8 x i8> %strided.vec, ptr %S
@@ -340,7 +205,7 @@ define void @shuffle_v64i8_to_v8i8_3(ptr %L, ptr %S) nounwind {
define void @shuffle_v64i8_to_v8i8_4(ptr %L, ptr %S) nounwind {
; AVX512-LABEL: shuffle_v64i8_to_v8i8_4:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpshufd {{.*#+}} zmm0 = mem[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512-NEXT: vpsrlq $32, (%rdi), %zmm0
; AVX512-NEXT: vpmovqb %zmm0, (%rsi)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
@@ -351,35 +216,12 @@ define void @shuffle_v64i8_to_v8i8_4(ptr %L, ptr %S) nounwind {
}
define void @shuffle_v64i8_to_v8i8_5(ptr %L, ptr %S) nounwind {
-; AVX512F-LABEL: shuffle_v64i8_to_v8i8_5:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpsrlq $40, (%rdi), %zmm0
-; AVX512F-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512F-NEXT: vmovq %xmm0, (%rsi)
-; AVX512F-NEXT: vzeroupper
-; AVX512F-NEXT: retq
-;
-; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_5:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpsrlq $40, (%rdi), %zmm0
-; AVX512VL-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512VL-NEXT: vmovq %xmm0, (%rsi)
-; AVX512VL-NEXT: vzeroupper
-; AVX512VL-NEXT: retq
-;
-; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_5:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpsrlq $40, (%rdi), %zmm0
-; AVX512BW-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
-;
-; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_5:
-; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrlq $40, (%rdi), %zmm0
-; AVX512BWVL-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BWVL-NEXT: vzeroupper
-; AVX512BWVL-NEXT: retq
+; AVX512-LABEL: shuffle_v64i8_to_v8i8_5:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpsrlq $40, (%rdi), %zmm0
+; AVX512-NEXT: vpmovqb %zmm0, (%rsi)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
%vec = load <64 x i8>, ptr %L
%strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61>
store <8 x i8> %strided.vec, ptr %S
@@ -387,39 +229,12 @@ define void @shuffle_v64i8_to_v8i8_5(ptr %L, ptr %S) nounwind {
}
define void @shuffle_v64i8_to_v8i8_6(ptr %L, ptr %S) nounwind {
-; AVX512F-LABEL: shuffle_v64i8_to_v8i8_6:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, 32(%rdi), %zmm0, %zmm0
-; AVX512F-NEXT: vpsrlq $48, %zmm0, %zmm0
-; AVX512F-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512F-NEXT: vmovq %xmm0, (%rsi)
-; AVX512F-NEXT: vzeroupper
-; AVX512F-NEXT: retq
-;
-; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_6:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512VL-NEXT: vinserti64x4 $1, 32(%rdi), %zmm0, %zmm0
-; AVX512VL-NEXT: vpsrlq $48, %zmm0, %zmm0
-; AVX512VL-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512VL-NEXT: vmovq %xmm0, (%rsi)
-; AVX512VL-NEXT: vzeroupper
-; AVX512VL-NEXT: retq
-;
-; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_6:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpsrlq $48, (%rdi), %zmm0
-; AVX512BW-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
-;
-; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_6:
-; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrlq $48, (%rdi), %zmm0
-; AVX512BWVL-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BWVL-NEXT: vzeroupper
-; AVX512BWVL-NEXT: retq
+; AVX512-LABEL: shuffle_v64i8_to_v8i8_6:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpsrlq $48, (%rdi), %zmm0
+; AVX512-NEXT: vpmovqb %zmm0, (%rsi)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
%vec = load <64 x i8>, ptr %L
%strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62>
store <8 x i8> %strided.vec, ptr %S
@@ -427,38 +242,17 @@ define void @shuffle_v64i8_to_v8i8_6(ptr %L, ptr %S) nounwind {
}
define void @shuffle_v64i8_to_v8i8_7(ptr %L, ptr %S) nounwind {
-; AVX512F-LABEL: shuffle_v64i8_to_v8i8_7:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpsrlq $56, (%rdi), %zmm0
-; AVX512F-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512F-NEXT: vmovq %xmm0, (%rsi)
-; AVX512F-NEXT: vzeroupper
-; AVX512F-NEXT: retq
-;
-; AVX512VL-LABEL: shuffle_v64i8_to_v8i8_7:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpsrlq $56, (%rdi), %zmm0
-; AVX512VL-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512VL-NEXT: vmovq %xmm0, (%rsi)
-; AVX512VL-NEXT: vzeroupper
-; AVX512VL-NEXT: retq
-;
-; AVX512BW-LABEL: shuffle_v64i8_to_v8i8_7:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpsrlq $56, (%rdi), %zmm0
-; AVX512BW-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
-;
-; AVX512BWVL-LABEL: shuffle_v64i8_to_v8i8_7:
-; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrlq $56, (%rdi), %zmm0
-; AVX512BWVL-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BWVL-NEXT: vzeroupper
-; AVX512BWVL-NEXT: retq
+; AVX512-LABEL: shuffle_v64i8_to_v8i8_7:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpsrlq $56, (%rdi), %zmm0
+; AVX512-NEXT: vpmovqb %zmm0, (%rsi)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
%vec = load <64 x i8>, ptr %L
%strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63>
store <8 x i8> %strided.vec, ptr %S
ret void
}
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; AVX512VL: {{.*}}
diff --git a/llvm/test/CodeGen/X86/shuffle-vs-trunc-512.ll b/llvm/test/CodeGen/X86/shuffle-vs-trunc-512.ll
index bbf9db65f35f..9a1d3ad7733a 100644
--- a/llvm/test/CodeGen/X86/shuffle-vs-trunc-512.ll
+++ b/llvm/test/CodeGen/X86/shuffle-vs-trunc-512.ll
@@ -274,53 +274,12 @@ define void @trunc_v8i64_to_v8i8(ptr %L, ptr %S) nounwind {
}
define <16 x i8> @trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_61(<64 x i8> %x) {
-; AVX512F-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_61:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpsrlw $8, %ymm0, %ymm1
-; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; AVX512F-NEXT: vpsrlw $8, %ymm0, %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
-; AVX512F-NEXT: vpmovdb %zmm0, %xmm0
-; AVX512F-NEXT: vzeroupper
-; AVX512F-NEXT: retq
-;
-; AVX512VL-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_61:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpsrlw $8, %ymm0, %ymm1
-; AVX512VL-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; AVX512VL-NEXT: vpsrlw $8, %ymm0, %ymm0
-; AVX512VL-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
-; AVX512VL-NEXT: vpmovdb %zmm0, %xmm0
-; AVX512VL-NEXT: vzeroupper
-; AVX512VL-NEXT: retq
-;
-; AVX512BW-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_61:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpsrlw $8, %zmm0, %zmm0
-; AVX512BW-NEXT: vpmovdb %zmm0, %xmm0
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
-;
-; AVX512BWVL-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_61:
-; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vpsrlw $8, %zmm0, %zmm0
-; AVX512BWVL-NEXT: vpmovdb %zmm0, %xmm0
-; AVX512BWVL-NEXT: vzeroupper
-; AVX512BWVL-NEXT: retq
-;
-; AVX512VBMI-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_61:
-; AVX512VBMI: # %bb.0:
-; AVX512VBMI-NEXT: vpsrlw $8, %zmm0, %zmm0
-; AVX512VBMI-NEXT: vpmovdb %zmm0, %xmm0
-; AVX512VBMI-NEXT: vzeroupper
-; AVX512VBMI-NEXT: retq
-;
-; AVX512VBMIVL-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_61:
-; AVX512VBMIVL: # %bb.0:
-; AVX512VBMIVL-NEXT: vpsrlw $8, %zmm0, %zmm0
-; AVX512VBMIVL-NEXT: vpmovdb %zmm0, %xmm0
-; AVX512VBMIVL-NEXT: vzeroupper
-; AVX512VBMIVL-NEXT: retq
+; AVX512-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_61:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpsrld $8, %zmm0, %zmm0
+; AVX512-NEXT: vpmovdb %zmm0, %xmm0
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
%res = shufflevector <64 x i8> %x, <64 x i8> %x, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
ret <16 x i8> %res
}
@@ -333,7 +292,7 @@ define <16 x i8> @trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_
; AVX512F-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,1,5,9,14,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; AVX512F-NEXT: vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT: vpsrld $8, %ymm0, %ymm0
; AVX512F-NEXT: vpmovdb %zmm0, %xmm0
; AVX512F-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512F-NEXT: vzeroupper
@@ -346,7 +305,7 @@ define <16 x i8> @trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_
; AVX512VL-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,1,5,9,14,u,u,u,u,u,u,u,u]
; AVX512VL-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u]
; AVX512VL-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; AVX512VL-NEXT: vpsrlw $8, %ymm0, %ymm0
+; AVX512VL-NEXT: vpsrld $8, %ymm0, %ymm0
; AVX512VL-NEXT: vpmovdb %ymm0, %xmm0
; AVX512VL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512VL-NEXT: vzeroupper
@@ -359,7 +318,7 @@ define <16 x i8> @trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_
; AVX512BW-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,1,5,9,14,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; AVX512BW-NEXT: vpsrlw $8, %ymm0, %ymm0
+; AVX512BW-NEXT: vpsrld $8, %ymm0, %ymm0
; AVX512BW-NEXT: vpmovdb %zmm0, %xmm0
; AVX512BW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512BW-NEXT: vzeroupper
@@ -372,7 +331,7 @@ define <16 x i8> @trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_
; AVX512BWVL-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,1,5,9,14,u,u,u,u,u,u,u,u]
; AVX512BWVL-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u]
; AVX512BWVL-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; AVX512BWVL-NEXT: vpsrlw $8, %ymm0, %ymm0
+; AVX512BWVL-NEXT: vpsrld $8, %ymm0, %ymm0
; AVX512BWVL-NEXT: vpmovdb %ymm0, %xmm0
; AVX512BWVL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512BWVL-NEXT: vzeroupper
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll
index dcaac8d2aa33..2c8d6573cdd1 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll
@@ -251,8 +251,8 @@ define void @load_i16_stride4_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512-LABEL: load_i16_stride4_vf4:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512-NEXT: vpsrld $16, %ymm0, %ymm1
-; AVX512-NEXT: vpshufd {{.*#+}} ymm2 = ymm0[1,1,3,3,5,5,7,7]
+; AVX512-NEXT: vpsrlq $16, %ymm0, %ymm1
+; AVX512-NEXT: vpsrlq $32, %ymm0, %ymm2
; AVX512-NEXT: vpsrlq $48, %ymm0, %ymm3
; AVX512-NEXT: vpmovqw %ymm0, (%rsi)
; AVX512-NEXT: vpmovqw %ymm1, (%rdx)
@@ -516,31 +516,18 @@ define void @load_i16_stride4_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
;
-; AVX512F-LABEL: load_i16_stride4_vf8:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpsrld $16, (%rdi), %zmm0
-; AVX512F-NEXT: vmovdqa64 (%rdi), %zmm1
-; AVX512F-NEXT: vpshufd {{.*#+}} zmm2 = zmm1[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
-; AVX512F-NEXT: vpsrlq $48, (%rdi), %zmm3
-; AVX512F-NEXT: vpmovqw %zmm1, (%rsi)
-; AVX512F-NEXT: vpmovqw %zmm0, (%rdx)
-; AVX512F-NEXT: vpmovqw %zmm2, (%rcx)
-; AVX512F-NEXT: vpmovqw %zmm3, (%r8)
-; AVX512F-NEXT: vzeroupper
-; AVX512F-NEXT: retq
-;
-; AVX512BW-LABEL: load_i16_stride4_vf8:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
-; AVX512BW-NEXT: vpsrld $16, %zmm0, %zmm1
-; AVX512BW-NEXT: vpshufd {{.*#+}} zmm2 = zmm0[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
-; AVX512BW-NEXT: vpsrlq $48, %zmm0, %zmm3
-; AVX512BW-NEXT: vpmovqw %zmm0, (%rsi)
-; AVX512BW-NEXT: vpmovqw %zmm1, (%rdx)
-; AVX512BW-NEXT: vpmovqw %zmm2, (%rcx)
-; AVX512BW-NEXT: vpmovqw %zmm3, (%r8)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
+; AVX512-LABEL: load_i16_stride4_vf8:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0
+; AVX512-NEXT: vpsrlq $16, %zmm0, %zmm1
+; AVX512-NEXT: vpsrlq $32, %zmm0, %zmm2
+; AVX512-NEXT: vpsrlq $48, %zmm0, %zmm3
+; AVX512-NEXT: vpmovqw %zmm0, (%rsi)
+; AVX512-NEXT: vpmovqw %zmm1, (%rdx)
+; AVX512-NEXT: vpmovqw %zmm2, (%rcx)
+; AVX512-NEXT: vpmovqw %zmm3, (%r8)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
%wide.vec = load <32 x i16>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
%strided.vec1 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
@@ -1059,7 +1046,7 @@ define void @load_i16_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
-; AVX512F-SLOW-NEXT: vpsrld $16, %zmm3, %zmm5
+; AVX512F-SLOW-NEXT: vpsrlq $16, %zmm3, %zmm5
; AVX512F-SLOW-NEXT: vpmovqw %zmm5, %xmm5
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
@@ -1075,7 +1062,7 @@ define void @load_i16_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4,5],ymm5[6,7]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} zmm8 = zmm3[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-SLOW-NEXT: vpsrlq $32, %zmm3, %zmm8
; AVX512F-SLOW-NEXT: vpmovqw %zmm8, %xmm8
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3],ymm5[4,5,6,7]
; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
@@ -1118,7 +1105,7 @@ define void @load_i16_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = <2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,u,u,u,u,u,u,u,u>
; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm1, %ymm1
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX512F-FAST-NEXT: vpsrld $16, %zmm4, %zmm3
+; AVX512F-FAST-NEXT: vpsrlq $16, %zmm4, %zmm3
; AVX512F-FAST-NEXT: vpmovqw %zmm3, %xmm3
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [1,3,2,3,1,3,5,7]
@@ -1127,7 +1114,7 @@ define void @load_i16_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpermd %ymm6, %ymm3, %ymm3
; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm3, %ymm5
; AVX512F-FAST-NEXT: vpermt2d %ymm0, %ymm8, %ymm5
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} zmm0 = zmm4[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-FAST-NEXT: vpsrlq $32, %zmm4, %zmm0
; AVX512F-FAST-NEXT: vpmovqw %zmm0, %xmm0
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm2, %ymm2
@@ -2321,7 +2308,7 @@ define void @load_i16_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm15[0],xmm9[0],xmm15[1],xmm9[1]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3,4,5],ymm5[6,7]
-; AVX512F-SLOW-NEXT: vpsrld $16, %zmm1, %zmm9
+; AVX512F-SLOW-NEXT: vpsrlq $16, %zmm1, %zmm9
; AVX512F-SLOW-NEXT: vpmovqw %zmm9, %xmm9
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3],ymm5[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm0, %zmm5
@@ -2336,7 +2323,7 @@ define void @load_i16_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6,7]
-; AVX512F-SLOW-NEXT: vpsrld $16, %zmm0, %zmm9
+; AVX512F-SLOW-NEXT: vpsrlq $16, %zmm0, %zmm9
; AVX512F-SLOW-NEXT: vpmovqw %zmm9, %xmm9
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm5 = zmm8[0,1,2,3],zmm5[4,5,6,7]
@@ -2353,7 +2340,7 @@ define void @load_i16_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} zmm13 = zmm1[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-SLOW-NEXT: vpsrlq $32, %zmm1, %zmm13
; AVX512F-SLOW-NEXT: vpmovqw %zmm13, %xmm13
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3],ymm12[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm12, %zmm0, %zmm12
@@ -2370,7 +2357,7 @@ define void @load_i16_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vpunpckldq {{.*#+}} xmm14 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} zmm14 = zmm0[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-SLOW-NEXT: vpsrlq $32, %zmm0, %zmm14
; AVX512F-SLOW-NEXT: vpmovqw %zmm14, %xmm14
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm12 = zmm13[0,1,2,3],zmm12[4,5,6,7]
@@ -2440,14 +2427,14 @@ define void @load_i16_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = <2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,u,u,u,u,u,u,u,u>
; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm9, %ymm9
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm13[6,7]
-; AVX512F-FAST-NEXT: vpsrld $16, %zmm1, %zmm13
+; AVX512F-FAST-NEXT: vpsrlq $16, %zmm1, %zmm13
; AVX512F-FAST-NEXT: vpmovqw %zmm13, %xmm13
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm13[0,1,2,3],ymm9[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm0, %zmm9
; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm12, %ymm12
; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm15, %ymm13
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
-; AVX512F-FAST-NEXT: vpsrld $16, %zmm0, %zmm13
+; AVX512F-FAST-NEXT: vpsrlq $16, %zmm0, %zmm13
; AVX512F-FAST-NEXT: vpmovqw %zmm13, %xmm13
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3],ymm12[4,5,6,7]
; AVX512F-FAST-NEXT: vshufi64x2 {{.*#+}} zmm9 = zmm12[0,1,2,3],zmm9[4,5,6,7]
@@ -2457,7 +2444,7 @@ define void @load_i16_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpermd %ymm8, %ymm12, %ymm8
; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm8, %ymm15
; AVX512F-FAST-NEXT: vpermt2d %ymm13, %ymm11, %ymm15
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} zmm13 = zmm1[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-FAST-NEXT: vpsrlq $32, %zmm1, %zmm13
; AVX512F-FAST-NEXT: vpmovqw %zmm13, %xmm13
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm15[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm13, %zmm0, %zmm13
@@ -2466,7 +2453,7 @@ define void @load_i16_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpermd %ymm14, %ymm12, %ymm12
; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm12, %ymm3
; AVX512F-FAST-NEXT: vpermt2d %ymm2, %ymm11, %ymm3
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} zmm2 = zmm0[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-FAST-NEXT: vpsrlq $32, %zmm0, %zmm2
; AVX512F-FAST-NEXT: vpmovqw %zmm2, %xmm2
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX512F-FAST-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm13[4,5,6,7]
@@ -5104,13 +5091,13 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT: vinserti128 $1, (%rsp), %ymm0, %ymm13 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
-; AVX512F-SLOW-NEXT: vpsrld $16, %zmm0, %zmm13
+; AVX512F-SLOW-NEXT: vpsrlq $16, %zmm0, %zmm13
; AVX512F-SLOW-NEXT: vpmovqw %zmm13, %xmm13
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3],ymm12[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT: vinserti32x4 $1, %xmm28, %ymm0, %ymm14
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
-; AVX512F-SLOW-NEXT: vpsrld $16, %zmm1, %zmm14
+; AVX512F-SLOW-NEXT: vpsrlq $16, %zmm1, %zmm14
; AVX512F-SLOW-NEXT: vpmovqw %zmm14, %xmm14
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm12, %zmm0, %zmm12
@@ -5118,13 +5105,13 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
-; AVX512F-SLOW-NEXT: vpsrld $16, %zmm2, %zmm13
+; AVX512F-SLOW-NEXT: vpsrlq $16, %zmm2, %zmm13
; AVX512F-SLOW-NEXT: vpmovqw %zmm13, %xmm13
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3],ymm12[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT: vinserti32x4 $1, %xmm24, %ymm0, %ymm14
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
-; AVX512F-SLOW-NEXT: vpsrld $16, %zmm21, %zmm14
+; AVX512F-SLOW-NEXT: vpsrlq $16, %zmm21, %zmm14
; AVX512F-SLOW-NEXT: vpmovqw %zmm14, %xmm14
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm12, %zmm0, %zmm12
@@ -5136,7 +5123,7 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT: vinserti32x4 $1, %xmm30, %ymm0, %ymm12
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm14[6,7]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} zmm14 = zmm0[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-SLOW-NEXT: vpsrlq $32, %zmm0, %zmm14
; AVX512F-SLOW-NEXT: vpmovqw %zmm14, %xmm14
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm14[0,1,2,3],ymm12[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm12, %zmm0, %zmm12
@@ -5147,7 +5134,7 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vinserti32x4 $1, %xmm23, %ymm0, %ymm14
; AVX512F-SLOW-NEXT: vinserti32x4 $1, %xmm27, %ymm0, %ymm10
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm14[6,7]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} zmm14 = zmm1[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-SLOW-NEXT: vpsrlq $32, %zmm1, %zmm14
; AVX512F-SLOW-NEXT: vpmovqw %zmm14, %xmm14
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm14[0,1,2,3],ymm10[4,5,6,7]
; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
@@ -5158,7 +5145,7 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vinserti32x4 $1, %xmm18, %ymm0, %ymm14
; AVX512F-SLOW-NEXT: vinserti32x4 $1, %xmm17, %ymm0, %ymm8
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm14[6,7]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} zmm14 = zmm2[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-SLOW-NEXT: vpsrlq $32, %zmm2, %zmm14
; AVX512F-SLOW-NEXT: vpmovqw %zmm14, %xmm14
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm14[0,1,2,3],ymm8[4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm8, %zmm0, %zmm8
@@ -5169,7 +5156,7 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vinserti32x4 $1, %xmm16, %ymm0, %ymm14
; AVX512F-SLOW-NEXT: vinserti32x4 $1, %xmm20, %ymm0, %ymm5
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm14[6,7]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} zmm14 = zmm21[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-SLOW-NEXT: vpsrlq $32, %zmm21, %zmm14
; AVX512F-SLOW-NEXT: vpmovqw %zmm14, %xmm14
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm14[0,1,2,3],ymm5[4,5,6,7]
; AVX512F-SLOW-NEXT: vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm8[4,5,6,7]
@@ -5283,28 +5270,28 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = <2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,u,u,u,u,u,u,u,u>
; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm11, %ymm11
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm14[6,7]
-; AVX512F-FAST-NEXT: vpsrld $16, %zmm4, %zmm14
+; AVX512F-FAST-NEXT: vpsrlq $16, %zmm4, %zmm14
; AVX512F-FAST-NEXT: vpmovqw %zmm14, %xmm14
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm14[0,1,2,3],ymm11[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm0, %zmm11
; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm3, %ymm3
; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm8, %ymm8
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1,2,3,4,5],ymm3[6,7]
-; AVX512F-FAST-NEXT: vpsrld $16, %zmm30, %zmm8
+; AVX512F-FAST-NEXT: vpsrlq $16, %zmm30, %zmm8
; AVX512F-FAST-NEXT: vpmovqw %zmm8, %xmm8
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1,2,3],ymm3[4,5,6,7]
; AVX512F-FAST-NEXT: vshufi64x2 {{.*#+}} zmm29 = zmm3[0,1,2,3],zmm11[4,5,6,7]
; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm0, %ymm0
; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm12, %ymm3
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX512F-FAST-NEXT: vpsrld $16, %zmm26, %zmm3
+; AVX512F-FAST-NEXT: vpsrlq $16, %zmm26, %zmm3
; AVX512F-FAST-NEXT: vpmovqw %zmm3, %xmm3
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm13, %ymm3
; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm1, %ymm1
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX512F-FAST-NEXT: vpsrld $16, %zmm23, %zmm3
+; AVX512F-FAST-NEXT: vpsrlq $16, %zmm23, %zmm3
; AVX512F-FAST-NEXT: vpmovqw %zmm3, %xmm3
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-FAST-NEXT: vshufi64x2 {{.*#+}} zmm19 = zmm1[0,1,2,3],zmm0[4,5,6,7]
@@ -5314,7 +5301,7 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpermd %ymm25, %ymm15, %ymm8
; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm8, %ymm1
; AVX512F-FAST-NEXT: vpermt2d %ymm0, %ymm7, %ymm1
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} zmm0 = zmm4[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-FAST-NEXT: vpsrlq $32, %zmm4, %zmm0
; AVX512F-FAST-NEXT: vpmovqw %zmm0, %xmm0
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm1
@@ -5323,7 +5310,7 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpermd %ymm28, %ymm15, %ymm12
; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm12, %ymm14
; AVX512F-FAST-NEXT: vpermt2d %ymm13, %ymm7, %ymm14
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} zmm13 = zmm30[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-FAST-NEXT: vpsrlq $32, %zmm30, %zmm13
; AVX512F-FAST-NEXT: vpmovqw %zmm13, %xmm13
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm14[4,5,6,7]
; AVX512F-FAST-NEXT: vshufi64x2 {{.*#+}} zmm24 = zmm13[0,1,2,3],zmm1[4,5,6,7]
@@ -5332,7 +5319,7 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpermd %ymm17, %ymm15, %ymm14
; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm14, %ymm11
; AVX512F-FAST-NEXT: vpermt2d %ymm1, %ymm7, %ymm11
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} zmm1 = zmm26[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-FAST-NEXT: vpsrlq $32, %zmm26, %zmm1
; AVX512F-FAST-NEXT: vpmovqw %zmm1, %xmm1
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm11[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm11
@@ -5341,7 +5328,7 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpermd %ymm20, %ymm15, %ymm5
; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm5, %ymm6
; AVX512F-FAST-NEXT: vpermt2d %ymm2, %ymm7, %ymm6
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} zmm2 = zmm23[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-FAST-NEXT: vpsrlq $32, %zmm23, %zmm2
; AVX512F-FAST-NEXT: vpmovqw %zmm2, %xmm2
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5,6,7]
; AVX512F-FAST-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm11[4,5,6,7]
@@ -5466,6 +5453,7 @@ define void @load_i16_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW: {{.*}}
; AVX512DQBW-FAST: {{.*}}
; AVX512DQBW-SLOW: {{.*}}
+; AVX512F: {{.*}}
; AVX512F-ONLY-FAST: {{.*}}
; AVX512F-ONLY-SLOW: {{.*}}
; FALLBACK0: {{.*}}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
index 7cb62c156a0c..042c6d2cc02b 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
@@ -245,7 +245,7 @@ define void @load_i8_stride4_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512-LABEL: load_i8_stride4_vf8:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512-NEXT: vpsrlw $8, %ymm0, %ymm1
+; AVX512-NEXT: vpsrld $8, %ymm0, %ymm1
; AVX512-NEXT: vpsrld $16, %ymm0, %ymm2
; AVX512-NEXT: vpsrld $24, %ymm0, %ymm3
; AVX512-NEXT: vpmovdb %ymm0, (%rsi)
@@ -475,35 +475,18 @@ define void @load_i8_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-NEXT: vmovdqa %xmm0, (%r8)
; AVX2-ONLY-NEXT: retq
;
-; AVX512F-LABEL: load_i8_stride4_vf16:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT: vmovdqa 32(%rdi), %ymm1
-; AVX512F-NEXT: vpsrlw $8, %ymm0, %ymm0
-; AVX512F-NEXT: vpsrlw $8, %ymm1, %ymm1
-; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512F-NEXT: vmovdqa64 (%rdi), %zmm1
-; AVX512F-NEXT: vpsrld $16, %zmm1, %zmm2
-; AVX512F-NEXT: vpsrld $24, %zmm1, %zmm3
-; AVX512F-NEXT: vpmovdb %zmm1, (%rsi)
-; AVX512F-NEXT: vpmovdb %zmm0, (%rdx)
-; AVX512F-NEXT: vpmovdb %zmm2, (%rcx)
-; AVX512F-NEXT: vpmovdb %zmm3, (%r8)
-; AVX512F-NEXT: vzeroupper
-; AVX512F-NEXT: retq
-;
-; AVX512BW-LABEL: load_i8_stride4_vf16:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
-; AVX512BW-NEXT: vpsrlw $8, %zmm0, %zmm1
-; AVX512BW-NEXT: vpsrld $16, %zmm0, %zmm2
-; AVX512BW-NEXT: vpsrld $24, %zmm0, %zmm3
-; AVX512BW-NEXT: vpmovdb %zmm0, (%rsi)
-; AVX512BW-NEXT: vpmovdb %zmm1, (%rdx)
-; AVX512BW-NEXT: vpmovdb %zmm2, (%rcx)
-; AVX512BW-NEXT: vpmovdb %zmm3, (%r8)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
+; AVX512-LABEL: load_i8_stride4_vf16:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0
+; AVX512-NEXT: vpsrld $8, %zmm0, %zmm1
+; AVX512-NEXT: vpsrld $16, %zmm0, %zmm2
+; AVX512-NEXT: vpsrld $24, %zmm0, %zmm3
+; AVX512-NEXT: vpmovdb %zmm0, (%rsi)
+; AVX512-NEXT: vpmovdb %zmm1, (%rdx)
+; AVX512-NEXT: vpmovdb %zmm2, (%rcx)
+; AVX512-NEXT: vpmovdb %zmm3, (%r8)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
%wide.vec = load <64 x i8>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
%strided.vec1 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
@@ -935,45 +918,41 @@ define void @load_i8_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
;
; AVX512F-LABEL: load_i8_stride4_vf32:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm1 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-NEXT: vmovdqa (%rdi), %ymm2
-; AVX512F-NEXT: vmovdqa 32(%rdi), %ymm3
-; AVX512F-NEXT: vmovdqa 64(%rdi), %ymm0
-; AVX512F-NEXT: vmovdqa 96(%rdi), %ymm4
-; AVX512F-NEXT: vpshufb %ymm1, %ymm4, %ymm5
-; AVX512F-NEXT: vpshufb %ymm1, %ymm0, %ymm1
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm6 = [0,4,0,4,0,4,8,12]
-; AVX512F-NEXT: vpermt2d %ymm5, %ymm6, %ymm1
-; AVX512F-NEXT: vmovdqa64 (%rdi), %zmm5
-; AVX512F-NEXT: vpmovdb %zmm5, %xmm7
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm7 = <1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-NEXT: vpshufb %ymm7, %ymm4, %ymm8
-; AVX512F-NEXT: vpshufb %ymm7, %ymm0, %ymm7
-; AVX512F-NEXT: vpermt2d %ymm8, %ymm6, %ymm7
-; AVX512F-NEXT: vpsrlw $8, %ymm2, %ymm2
-; AVX512F-NEXT: vpsrlw $8, %ymm3, %ymm3
-; AVX512F-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovdb %zmm2, %xmm2
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm3 = <2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-NEXT: vpshufb %ymm3, %ymm4, %ymm7
-; AVX512F-NEXT: vpshufb %ymm3, %ymm0, %ymm3
-; AVX512F-NEXT: vpermt2d %ymm7, %ymm6, %ymm3
-; AVX512F-NEXT: vpsrld $16, %zmm5, %zmm7
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm0 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-NEXT: vmovdqa 96(%rdi), %ymm1
+; AVX512F-NEXT: vpshufb %ymm0, %ymm1, %ymm2
+; AVX512F-NEXT: vmovdqa 64(%rdi), %ymm3
+; AVX512F-NEXT: vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm4 = [0,4,0,4,0,4,8,12]
+; AVX512F-NEXT: vpermt2d %ymm2, %ymm4, %ymm0
+; AVX512F-NEXT: vmovdqa64 (%rdi), %zmm2
+; AVX512F-NEXT: vpmovdb %zmm2, %xmm5
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm5 = <1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-NEXT: vpshufb %ymm5, %ymm1, %ymm6
+; AVX512F-NEXT: vpshufb %ymm5, %ymm3, %ymm5
+; AVX512F-NEXT: vpermt2d %ymm6, %ymm4, %ymm5
+; AVX512F-NEXT: vpsrld $8, %zmm2, %zmm6
+; AVX512F-NEXT: vpmovdb %zmm6, %xmm6
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm6 = <2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-NEXT: vpshufb %ymm6, %ymm1, %ymm7
+; AVX512F-NEXT: vpshufb %ymm6, %ymm3, %ymm6
+; AVX512F-NEXT: vpermt2d %ymm7, %ymm4, %ymm6
+; AVX512F-NEXT: vpsrld $16, %zmm2, %zmm7
; AVX512F-NEXT: vpmovdb %zmm7, %xmm7
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX512F-NEXT: vmovdqa {{.*#+}} ymm7 = <3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-NEXT: vpshufb %ymm7, %ymm4, %ymm4
-; AVX512F-NEXT: vpshufb %ymm7, %ymm0, %ymm0
-; AVX512F-NEXT: vpermt2d %ymm4, %ymm6, %ymm0
-; AVX512F-NEXT: vpsrld $24, %zmm5, %zmm4
-; AVX512F-NEXT: vpmovdb %zmm4, %xmm4
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-NEXT: vmovdqa %ymm1, (%rsi)
-; AVX512F-NEXT: vmovdqa %ymm2, (%rdx)
-; AVX512F-NEXT: vmovdqa %ymm3, (%rcx)
-; AVX512F-NEXT: vmovdqa %ymm0, (%r8)
+; AVX512F-NEXT: vpshufb %ymm7, %ymm1, %ymm1
+; AVX512F-NEXT: vpshufb %ymm7, %ymm3, %ymm3
+; AVX512F-NEXT: vpermt2d %ymm1, %ymm4, %ymm3
+; AVX512F-NEXT: vpsrld $24, %zmm2, %zmm1
+; AVX512F-NEXT: vpmovdb %zmm1, %xmm1
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX512F-NEXT: vmovdqa %ymm0, (%rsi)
+; AVX512F-NEXT: vmovdqa %ymm5, (%rdx)
+; AVX512F-NEXT: vmovdqa %ymm6, (%rcx)
+; AVX512F-NEXT: vmovdqa %ymm1, (%r8)
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -1937,47 +1916,39 @@ define void @load_i8_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512F-NEXT: vmovdqa64 128(%rdi), %zmm2
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm6 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm7 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
; AVX512F-NEXT: vmovdqa 224(%rdi), %ymm3
-; AVX512F-NEXT: vpshufb %ymm6, %ymm3, %ymm5
+; AVX512F-NEXT: vpshufb %ymm7, %ymm3, %ymm5
; AVX512F-NEXT: vmovdqa 192(%rdi), %ymm4
-; AVX512F-NEXT: vpshufb %ymm6, %ymm4, %ymm7
+; AVX512F-NEXT: vpshufb %ymm7, %ymm4, %ymm6
; AVX512F-NEXT: vmovdqa {{.*#+}} ymm1 = [0,4,0,4,0,4,8,12]
-; AVX512F-NEXT: vpermt2d %ymm5, %ymm1, %ymm7
+; AVX512F-NEXT: vpermt2d %ymm5, %ymm1, %ymm6
; AVX512F-NEXT: vpmovdb %zmm2, %xmm5
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm7[4,5,6,7]
-; AVX512F-NEXT: vinserti64x4 $1, %ymm5, %zmm0, %zmm9
-; AVX512F-NEXT: vmovdqa (%rdi), %ymm10
-; AVX512F-NEXT: vmovdqa 32(%rdi), %ymm8
-; AVX512F-NEXT: vmovdqa 64(%rdi), %ymm5
-; AVX512F-NEXT: vmovdqa 96(%rdi), %ymm7
-; AVX512F-NEXT: vpshufb %ymm6, %ymm7, %ymm11
-; AVX512F-NEXT: vpshufb %ymm6, %ymm5, %ymm6
-; AVX512F-NEXT: vpermt2d %ymm11, %ymm1, %ymm6
-; AVX512F-NEXT: vpmovdb %zmm0, %xmm11
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm6 = ymm11[0,1,2,3],ymm6[4,5,6,7]
-; AVX512F-NEXT: vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm9[4,5,6,7]
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm9 = <1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-NEXT: vpshufb %ymm9, %ymm3, %ymm11
-; AVX512F-NEXT: vpshufb %ymm9, %ymm4, %ymm12
-; AVX512F-NEXT: vpermt2d %ymm11, %ymm1, %ymm12
-; AVX512F-NEXT: vmovdqa 128(%rdi), %ymm11
-; AVX512F-NEXT: vpsrlw $8, %ymm11, %ymm11
-; AVX512F-NEXT: vmovdqa 160(%rdi), %ymm13
-; AVX512F-NEXT: vpsrlw $8, %ymm13, %ymm13
-; AVX512F-NEXT: vinserti64x4 $1, %ymm13, %zmm11, %zmm11
-; AVX512F-NEXT: vpmovdb %zmm11, %xmm11
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
-; AVX512F-NEXT: vinserti64x4 $1, %ymm11, %zmm0, %zmm11
-; AVX512F-NEXT: vpshufb %ymm9, %ymm7, %ymm12
-; AVX512F-NEXT: vpshufb %ymm9, %ymm5, %ymm9
-; AVX512F-NEXT: vpermt2d %ymm12, %ymm1, %ymm9
-; AVX512F-NEXT: vpsrlw $8, %ymm10, %ymm10
-; AVX512F-NEXT: vpsrlw $8, %ymm8, %ymm8
-; AVX512F-NEXT: vinserti64x4 $1, %ymm8, %zmm10, %zmm8
-; AVX512F-NEXT: vpmovdb %zmm8, %xmm8
-; AVX512F-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7]
-; AVX512F-NEXT: vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm11[4,5,6,7]
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX512F-NEXT: vinserti64x4 $1, %ymm5, %zmm0, %zmm8
+; AVX512F-NEXT: vmovdqa 96(%rdi), %ymm5
+; AVX512F-NEXT: vpshufb %ymm7, %ymm5, %ymm9
+; AVX512F-NEXT: vmovdqa 64(%rdi), %ymm6
+; AVX512F-NEXT: vpshufb %ymm7, %ymm6, %ymm7
+; AVX512F-NEXT: vpermt2d %ymm9, %ymm1, %ymm7
+; AVX512F-NEXT: vpmovdb %zmm0, %xmm9
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
+; AVX512F-NEXT: vshufi64x2 {{.*#+}} zmm7 = zmm7[0,1,2,3],zmm8[4,5,6,7]
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm8 = <1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-NEXT: vpshufb %ymm8, %ymm3, %ymm9
+; AVX512F-NEXT: vpshufb %ymm8, %ymm4, %ymm10
+; AVX512F-NEXT: vpermt2d %ymm9, %ymm1, %ymm10
+; AVX512F-NEXT: vpsrld $8, %zmm2, %zmm9
+; AVX512F-NEXT: vpmovdb %zmm9, %xmm9
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
+; AVX512F-NEXT: vinserti64x4 $1, %ymm9, %zmm0, %zmm9
+; AVX512F-NEXT: vpshufb %ymm8, %ymm5, %ymm10
+; AVX512F-NEXT: vpshufb %ymm8, %ymm6, %ymm8
+; AVX512F-NEXT: vpermt2d %ymm10, %ymm1, %ymm8
+; AVX512F-NEXT: vpsrld $8, %zmm0, %zmm10
+; AVX512F-NEXT: vpmovdb %zmm10, %xmm10
+; AVX512F-NEXT: vpblendd {{.*#+}} ymm8 = ymm10[0,1,2,3],ymm8[4,5,6,7]
+; AVX512F-NEXT: vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm9[4,5,6,7]
; AVX512F-NEXT: vmovdqa {{.*#+}} ymm9 = <2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u>
; AVX512F-NEXT: vpshufb %ymm9, %ymm3, %ymm10
; AVX512F-NEXT: vpshufb %ymm9, %ymm4, %ymm11
@@ -1986,8 +1957,8 @@ define void @load_i8_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-NEXT: vpmovdb %zmm10, %xmm10
; AVX512F-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
; AVX512F-NEXT: vinserti64x4 $1, %ymm10, %zmm0, %zmm10
-; AVX512F-NEXT: vpshufb %ymm9, %ymm7, %ymm11
-; AVX512F-NEXT: vpshufb %ymm9, %ymm5, %ymm9
+; AVX512F-NEXT: vpshufb %ymm9, %ymm5, %ymm11
+; AVX512F-NEXT: vpshufb %ymm9, %ymm6, %ymm9
; AVX512F-NEXT: vpermt2d %ymm11, %ymm1, %ymm9
; AVX512F-NEXT: vpsrld $16, %zmm0, %zmm11
; AVX512F-NEXT: vpmovdb %zmm11, %xmm11
@@ -2001,14 +1972,14 @@ define void @load_i8_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-NEXT: vpmovdb %zmm2, %xmm2
; AVX512F-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm2
-; AVX512F-NEXT: vpshufb %ymm10, %ymm7, %ymm3
-; AVX512F-NEXT: vpshufb %ymm10, %ymm5, %ymm4
+; AVX512F-NEXT: vpshufb %ymm10, %ymm5, %ymm3
+; AVX512F-NEXT: vpshufb %ymm10, %ymm6, %ymm4
; AVX512F-NEXT: vpermt2d %ymm3, %ymm1, %ymm4
; AVX512F-NEXT: vpsrld $24, %zmm0, %zmm0
; AVX512F-NEXT: vpmovdb %zmm0, %xmm0
; AVX512F-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
; AVX512F-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
-; AVX512F-NEXT: vmovdqa64 %zmm6, (%rsi)
+; AVX512F-NEXT: vmovdqa64 %zmm7, (%rsi)
; AVX512F-NEXT: vmovdqa64 %zmm8, (%rdx)
; AVX512F-NEXT: vmovdqa64 %zmm9, (%rcx)
; AVX512F-NEXT: vmovdqa64 %zmm0, (%r8)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll
index 455b77c95cd6..abcd45362369 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll
@@ -290,10 +290,10 @@ define void @load_i8_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512-NEXT: movq {{[0-9]+}}(%rsp), %r10
; AVX512-NEXT: movq {{[0-9]+}}(%rsp), %r11
; AVX512-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512-NEXT: vpsrlw $8, %ymm0, %ymm1
-; AVX512-NEXT: vpsrld $16, %ymm0, %ymm2
-; AVX512-NEXT: vpsrld $24, %ymm0, %ymm3
-; AVX512-NEXT: vpshufd {{.*#+}} ymm4 = ymm0[1,1,3,3,5,5,7,7]
+; AVX512-NEXT: vpsrlq $8, %ymm0, %ymm1
+; AVX512-NEXT: vpsrlq $16, %ymm0, %ymm2
+; AVX512-NEXT: vpsrlq $24, %ymm0, %ymm3
+; AVX512-NEXT: vpsrlq $32, %ymm0, %ymm4
; AVX512-NEXT: vpsrlq $40, %ymm0, %ymm5
; AVX512-NEXT: vpsrlq $48, %ymm0, %ymm6
; AVX512-NEXT: vpsrlq $56, %ymm0, %ymm7
@@ -720,67 +720,29 @@ define void @load_i8_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-NEXT: vmovq %xmm0, (%rax)
; AVX2-ONLY-NEXT: retq
;
-; AVX512F-LABEL: load_i8_stride8_vf8:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-NEXT: movq {{[0-9]+}}(%rsp), %r10
-; AVX512F-NEXT: movq {{[0-9]+}}(%rsp), %r11
-; AVX512F-NEXT: vmovdqa64 (%rdi), %zmm0
-; AVX512F-NEXT: vpmovqb %zmm0, %xmm1
-; AVX512F-NEXT: vmovdqa (%rdi), %ymm2
-; AVX512F-NEXT: vmovdqa 32(%rdi), %ymm3
-; AVX512F-NEXT: vpsrlw $8, %ymm2, %ymm2
-; AVX512F-NEXT: vpsrlw $8, %ymm3, %ymm3
-; AVX512F-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovqb %zmm2, %xmm2
-; AVX512F-NEXT: vmovdqa64 (%rdi), %zmm3
-; AVX512F-NEXT: vpsrld $16, %zmm3, %zmm4
-; AVX512F-NEXT: vpmovqb %zmm4, %xmm4
-; AVX512F-NEXT: vpsrld $24, %zmm3, %zmm3
-; AVX512F-NEXT: vpmovqb %zmm3, %xmm3
-; AVX512F-NEXT: vpshufd {{.*#+}} zmm0 = zmm0[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
-; AVX512F-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512F-NEXT: vmovdqa64 (%rdi), %zmm5
-; AVX512F-NEXT: vpsrlq $40, %zmm5, %zmm6
-; AVX512F-NEXT: vpmovqb %zmm6, %xmm6
-; AVX512F-NEXT: vpsrlq $48, %zmm5, %zmm7
-; AVX512F-NEXT: vpmovqb %zmm7, %xmm7
-; AVX512F-NEXT: vpsrlq $56, %zmm5, %zmm5
-; AVX512F-NEXT: vpmovqb %zmm5, %xmm5
-; AVX512F-NEXT: vmovq %xmm1, (%rsi)
-; AVX512F-NEXT: vmovq %xmm2, (%rdx)
-; AVX512F-NEXT: vmovq %xmm4, (%rcx)
-; AVX512F-NEXT: vmovq %xmm3, (%r8)
-; AVX512F-NEXT: vmovq %xmm0, (%r9)
-; AVX512F-NEXT: vmovq %xmm6, (%r11)
-; AVX512F-NEXT: vmovq %xmm7, (%r10)
-; AVX512F-NEXT: vmovq %xmm5, (%rax)
-; AVX512F-NEXT: vzeroupper
-; AVX512F-NEXT: retq
-;
-; AVX512BW-LABEL: load_i8_stride8_vf8:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512BW-NEXT: movq {{[0-9]+}}(%rsp), %r10
-; AVX512BW-NEXT: movq {{[0-9]+}}(%rsp), %r11
-; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
-; AVX512BW-NEXT: vpsrlw $8, %zmm0, %zmm1
-; AVX512BW-NEXT: vpsrld $16, %zmm0, %zmm2
-; AVX512BW-NEXT: vpsrld $24, %zmm0, %zmm3
-; AVX512BW-NEXT: vpshufd {{.*#+}} zmm4 = zmm0[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
-; AVX512BW-NEXT: vpsrlq $40, %zmm0, %zmm5
-; AVX512BW-NEXT: vpsrlq $48, %zmm0, %zmm6
-; AVX512BW-NEXT: vpsrlq $56, %zmm0, %zmm7
-; AVX512BW-NEXT: vpmovqb %zmm0, (%rsi)
-; AVX512BW-NEXT: vpmovqb %zmm1, (%rdx)
-; AVX512BW-NEXT: vpmovqb %zmm2, (%rcx)
-; AVX512BW-NEXT: vpmovqb %zmm3, (%r8)
-; AVX512BW-NEXT: vpmovqb %zmm4, (%r9)
-; AVX512BW-NEXT: vpmovqb %zmm5, (%r11)
-; AVX512BW-NEXT: vpmovqb %zmm6, (%r10)
-; AVX512BW-NEXT: vpmovqb %zmm7, (%rax)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
+; AVX512-LABEL: load_i8_stride8_vf8:
+; AVX512: # %bb.0:
+; AVX512-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX512-NEXT: movq {{[0-9]+}}(%rsp), %r10
+; AVX512-NEXT: movq {{[0-9]+}}(%rsp), %r11
+; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0
+; AVX512-NEXT: vpsrlq $8, %zmm0, %zmm1
+; AVX512-NEXT: vpsrlq $16, %zmm0, %zmm2
+; AVX512-NEXT: vpsrlq $24, %zmm0, %zmm3
+; AVX512-NEXT: vpsrlq $32, %zmm0, %zmm4
+; AVX512-NEXT: vpsrlq $40, %zmm0, %zmm5
+; AVX512-NEXT: vpsrlq $48, %zmm0, %zmm6
+; AVX512-NEXT: vpsrlq $56, %zmm0, %zmm7
+; AVX512-NEXT: vpmovqb %zmm0, (%rsi)
+; AVX512-NEXT: vpmovqb %zmm1, (%rdx)
+; AVX512-NEXT: vpmovqb %zmm2, (%rcx)
+; AVX512-NEXT: vpmovqb %zmm3, (%r8)
+; AVX512-NEXT: vpmovqb %zmm4, (%r9)
+; AVX512-NEXT: vpmovqb %zmm5, (%r11)
+; AVX512-NEXT: vpmovqb %zmm6, (%r10)
+; AVX512-NEXT: vpmovqb %zmm7, (%rax)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
%wide.vec = load <64 x i8>, ptr %in.vec, align 64
%strided.vec0 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <8 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56>
%strided.vec1 = shufflevector <64 x i8> %wide.vec, <64 x i8> poison, <8 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57>
@@ -5289,18 +5251,19 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: vmovdqa 240(%rdi), %xmm2
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,u,u,u,u,0,8,u,u,u,u,u,u,u,u>
; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm2, %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa %xmm2, %xmm9
+; AVX512F-SLOW-NEXT: vmovdqa %xmm2, %xmm12
; AVX512F-SLOW-NEXT: vmovdqa 224(%rdi), %xmm2
; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm2, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa %xmm2, %xmm6
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-SLOW-NEXT: vmovdqa 208(%rdi), %xmm6
+; AVX512F-SLOW-NEXT: vmovdqa 208(%rdi), %xmm2
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,u,u,0,8,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm6, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa %xmm6, %xmm7
-; AVX512F-SLOW-NEXT: vmovdqa 192(%rdi), %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm6, %xmm8
-; AVX512F-SLOW-NEXT: vmovdqa %xmm6, %xmm14
+; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm2, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa %xmm2, %xmm7
+; AVX512F-SLOW-NEXT: vmovdqa 192(%rdi), %xmm2
+; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm2, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa %xmm2, %xmm15
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm8[0],xmm3[0],xmm8[1],xmm3[1],xmm8[2],xmm3[2],xmm8[3],xmm3[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
@@ -5312,239 +5275,239 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT: vmovdqa 112(%rdi), %xmm1
; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm1, %xmm8
-; AVX512F-SLOW-NEXT: vmovdqa %xmm1, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa 96(%rdi), %xmm1
-; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm1, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm19
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm18
+; AVX512F-SLOW-NEXT: vmovdqa 96(%rdi), %xmm9
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm9, %xmm4
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3]
-; AVX512F-SLOW-NEXT: vmovdqa 80(%rdi), %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm6, %xmm11
-; AVX512F-SLOW-NEXT: vmovdqa 64(%rdi), %xmm8
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm8, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa 80(%rdi), %xmm1
+; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm1, %xmm11
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm19
+; AVX512F-SLOW-NEXT: vmovdqa 64(%rdi), %xmm1
+; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm1, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm20
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm11[0],xmm5[1],xmm11[1],xmm5[2],xmm11[2],xmm5[3],xmm11[3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm5[0,1,2],xmm10[3]
; AVX512F-SLOW-NEXT: vpmovqb %zmm16, %xmm10
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm0, %ymm18
-; AVX512F-SLOW-NEXT: vmovdqa 160(%rdi), %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa 176(%rdi), %xmm11
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,u,u,u,u,1,9,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm9, %xmm0
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm9, %xmm20
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm2, %xmm12
-; AVX512F-SLOW-NEXT: vmovdqa %xmm2, %xmm10
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm12
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,u,u,1,9,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm7, %xmm13
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm7, %xmm21
-; AVX512F-SLOW-NEXT: vmovdqa %xmm14, %xmm9
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm14, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1],xmm14[2],xmm13[2],xmm14[3],xmm13[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5,6],ymm12[7]
-; AVX512F-SLOW-NEXT: vmovdqa (%rdi), %ymm13
-; AVX512F-SLOW-NEXT: vmovdqa 32(%rdi), %ymm14
-; AVX512F-SLOW-NEXT: vpsrlw $8, %ymm13, %ymm13
-; AVX512F-SLOW-NEXT: vpsrlw $8, %ymm14, %ymm14
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm14, %zmm13, %zmm17
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm13 = <u,u,1,9,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm13, %xmm11, %xmm14
-; AVX512F-SLOW-NEXT: vpshufb %xmm13, %xmm1, %xmm13
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm22
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm13[0],xmm14[0],xmm13[1],xmm14[1],xmm13[2],xmm14[2],xmm13[3],xmm14[3]
-; AVX512F-SLOW-NEXT: vmovdqa 128(%rdi), %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa 144(%rdi), %xmm7
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm0, %ymm17
+; AVX512F-SLOW-NEXT: vmovdqa 160(%rdi), %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa 176(%rdi), %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,u,u,u,u,1,9,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vmovdqa %xmm12, %xmm4
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm12, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm6, %xmm12
+; AVX512F-SLOW-NEXT: vmovdqa %xmm6, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm12[0],xmm5[0],xmm12[1],xmm5[1],xmm12[2],xmm5[2],xmm12[3],xmm5[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm14 = <u,u,u,u,1,9,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm7, %xmm12
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm15, %xmm13
+; AVX512F-SLOW-NEXT: vmovdqa %xmm15, %xmm10
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm12[0,1,2,3,4,5,6],ymm5[7]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = <u,u,1,9,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm12, %xmm1, %xmm13
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm24
+; AVX512F-SLOW-NEXT: vpshufb %xmm12, %xmm0, %xmm12
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm0, %xmm21
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm15 = xmm12[0],xmm13[0],xmm12[1],xmm13[1],xmm12[2],xmm13[2],xmm12[3],xmm13[3]
+; AVX512F-SLOW-NEXT: vmovdqa 128(%rdi), %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa 144(%rdi), %xmm6
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm7, %xmm15
-; AVX512F-SLOW-NEXT: vmovdqa %xmm7, %xmm14
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm4, %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa %xmm4, %xmm13
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1],xmm1[2],xmm15[2],xmm1[3],xmm15[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa %xmm6, %xmm13
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm3, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa %xmm3, %xmm12
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm15, %ymm0, %ymm1
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm3, %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm4
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm4, %xmm5
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm6, %xmm5
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm8, %xmm2
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm18, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm1
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm2
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm6, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm11
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm11, %xmm5
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
-; AVX512F-SLOW-NEXT: vpmovqb %zmm17, %xmm2
+; AVX512F-SLOW-NEXT: vpsrlq $8, %zmm16, %zmm2
+; AVX512F-SLOW-NEXT: vpmovqb %zmm2, %xmm2
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm0, %ymm17
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm0, %ymm18
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,u,u,2,10,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm7
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm7, %xmm1
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm10, %xmm2
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm1
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm8, %xmm2
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,u,u,2,10,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm12
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm12, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm20
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm12
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm9, %xmm21
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm12[0],xmm5[0],xmm12[1],xmm5[1],xmm12[2],xmm5[2],xmm12[3],xmm5[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm7, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm14
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm10, %xmm23
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,2,10,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm11, %xmm12
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm9
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm9, %xmm5
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = <2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm12, %xmm14, %xmm15
-; AVX512F-SLOW-NEXT: vpshufb %xmm12, %xmm13, %xmm12
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm12[0],xmm15[0],xmm12[1],xmm15[1],xmm12[2],xmm15[2],xmm12[3],xmm15[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm10
+; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm10, %xmm14
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm10
+; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm10, %xmm5
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm14[0],xmm5[1],xmm14[1],xmm5[2],xmm14[2],xmm5[3],xmm14[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm14 = <2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm13, %xmm15
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm12, %xmm14
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1],xmm14[2],xmm15[2],xmm14[3],xmm15[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm12[0,1,2,3,4],ymm5[5],ymm12[6,7]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm5[5],ymm14[6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm3, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm3, %xmm22
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm0
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm24
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm3, %xmm20
+; AVX512F-SLOW-NEXT: vmovdqa %xmm9, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm9, %xmm0
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; AVX512F-SLOW-NEXT: vmovdqa %xmm6, %xmm3
; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm6, %xmm5
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm8, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa %xmm8, %xmm6
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm22
+; AVX512F-SLOW-NEXT: vmovdqa %xmm11, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm11, %xmm2
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3]
-; AVX512F-SLOW-NEXT: vpsrld $16, %zmm16, %zmm2
+; AVX512F-SLOW-NEXT: vpsrlq $16, %zmm16, %zmm2
; AVX512F-SLOW-NEXT: vpmovqb %zmm2, %xmm2
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-SLOW-NEXT: vmovdqa64 %ymm0, %ymm19
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,u,u,3,11,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa %xmm7, %xmm4
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm7, %xmm1
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm10, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm10, %xmm25
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm4, %xmm21
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm8, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm8, %xmm25
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,u,u,3,11,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm7
; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm7, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm8
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm8, %xmm12
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm12[0],xmm5[0],xmm12[1],xmm5[1],xmm12[2],xmm5[2],xmm12[3],xmm5[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm9
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm14
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,3,11,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm11, %xmm12
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm11, %xmm23
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm9, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa %xmm9, %xmm11
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = <3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm12, %xmm14, %xmm15
-; AVX512F-SLOW-NEXT: vpshufb %xmm12, %xmm13, %xmm12
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm12[0],xmm15[0],xmm12[1],xmm15[1],xmm12[2],xmm15[2],xmm12[3],xmm15[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm11
+; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm11, %xmm14
+; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm10, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm10, %xmm26
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm14[0],xmm5[1],xmm14[1],xmm5[2],xmm14[2],xmm5[3],xmm14[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm14 = <3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm13, %xmm15
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm12, %xmm14
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1],xmm14[2],xmm15[2],xmm14[3],xmm15[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm12[0,1,2,3,4],ymm5[5],ymm12[6,7]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm14[0,1,2,3,4],ymm5[5],ymm14[6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm10
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm10, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm9
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm9, %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm4
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa %xmm3, %xmm15
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm3, %xmm0
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm3, %xmm21
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm10
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm5
; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm6, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm6, %xmm24
+; AVX512F-SLOW-NEXT: vmovdqa %xmm6, %xmm8
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3]
-; AVX512F-SLOW-NEXT: vpsrld $24, %zmm16, %zmm2
+; AVX512F-SLOW-NEXT: vpsrlq $24, %zmm16, %zmm2
; AVX512F-SLOW-NEXT: vpmovqb %zmm2, %xmm2
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-SLOW-NEXT: vmovdqa64 %ymm0, %ymm20
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,u,u,u,u,4,12,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm4, %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm6, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm6, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm0
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm0, %xmm2
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,u,u,4,12,u,u,u,u,u,u,u,u,u,u>
; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm7, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm7, %xmm25
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm8, %xmm12
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm12[0],xmm5[0],xmm12[1],xmm5[1],xmm12[2],xmm5[2],xmm12[3],xmm5[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm14
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,4,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm7
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm7, %xmm12
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm11, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm11, %xmm23
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = <4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm12, %xmm14, %xmm0
-; AVX512F-SLOW-NEXT: vmovdqa %xmm14, %xmm15
-; AVX512F-SLOW-NEXT: vpshufb %xmm12, %xmm13, %xmm12
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm13, %xmm26
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm11, %xmm14
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm0
+; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm0, %xmm5
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm14[0],xmm5[1],xmm14[1],xmm5[2],xmm14[2],xmm5[3],xmm14[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm14 = <4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm13, %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm13, %xmm28
+; AVX512F-SLOW-NEXT: vpshufb %xmm14, %xmm12, %xmm14
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm12, %xmm24
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm14[0],xmm0[0],xmm14[1],xmm0[1],xmm14[2],xmm0[2],xmm14[3],xmm0[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5],ymm0[6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm11
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm11, %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa %xmm9, %xmm12
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm9, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm4, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm4, %xmm22
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm15, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm15, %xmm23
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm10
; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm9
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm2
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm8, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa %xmm8, %xmm14
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} zmm2 = zmm16[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-SLOW-NEXT: vpsrlq $32, %zmm16, %zmm2
; AVX512F-SLOW-NEXT: vpmovqb %zmm2, %xmm2
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT: vmovdqa64 %ymm0, %ymm21
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm13 = <u,u,u,u,u,u,5,13,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm13, %xmm4, %xmm1
-; AVX512F-SLOW-NEXT: vpshufb %xmm13, %xmm6, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa %xmm6, %xmm14
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = <u,u,u,u,u,u,5,13,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm12, %xmm6, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa %xmm6, %xmm13
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm4
+; AVX512F-SLOW-NEXT: vpshufb %xmm12, %xmm4, %xmm2
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,u,u,5,13,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm0
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm0, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm8, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm8, %xmm24
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm7, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm7, %xmm25
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa %xmm9, %xmm15
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,5,13,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm7, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm11, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm11, %xmm27
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm6
; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm6, %xmm3
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <5,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm15, %xmm0
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm11
+; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm11, %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm8
; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm8, %xmm5
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5],ymm0[6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm13, %xmm11, %xmm1
-; AVX512F-SLOW-NEXT: vpshufb %xmm13, %xmm12, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm12, %xmm5, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm7
+; AVX512F-SLOW-NEXT: vpshufb %xmm12, %xmm7, %xmm3
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm10, %xmm23
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm14, %xmm2
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
; AVX512F-SLOW-NEXT: vpsrlq $40, %zmm16, %zmm2
@@ -5552,39 +5515,43 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT: vmovdqa64 %ymm0, %ymm22
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,u,u,u,u,6,14,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm4, %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm4, %xmm23
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm14, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm14, %xmm27
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm9 = <u,u,u,u,u,u,6,14,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm9, %xmm13, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa %xmm13, %xmm12
+; AVX512F-SLOW-NEXT: vpshufb %xmm9, %xmm4, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa %xmm4, %xmm13
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,u,u,6,14,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm8
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm8, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm14
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm14, %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm10
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm15, %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm15, %xmm25
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,6,14,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm7, %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm27, %xmm0
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm0, %xmm4
; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm6, %xmm3
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm4 = <6,14,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm15, %xmm0
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm13
-; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm13, %xmm4
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm11, %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm15
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm8, %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm11
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5],ymm0[6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm11, %xmm1
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm12, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm9, %xmm5, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm24
+; AVX512F-SLOW-NEXT: vpshufb %xmm9, %xmm7, %xmm3
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm8
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm8, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm14, %xmm2
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
; AVX512F-SLOW-NEXT: vpsrlq $48, %zmm16, %zmm2
@@ -5592,43 +5559,44 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,u,u,7,15,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm2
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm2, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm27, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm3, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm12, %xmm2
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm13, %xmm3
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,u,u,7,15,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm8, %xmm4
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm14, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm10, %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm5, %xmm5
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,7,15,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm7, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm27, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm5, %xmm5
; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm6, %xmm4
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <7,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm15, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm13, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm11, %xmm5
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5],ymm5[6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm11, %xmm4
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm12, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm4
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm4, %xmm4
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm7, %xmm1
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm10, %xmm4
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm9, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm8, %xmm4
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm14, %xmm3
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
; AVX512F-SLOW-NEXT: vpsrlq $56, %zmm16, %zmm3
; AVX512F-SLOW-NEXT: vpmovqb %zmm3, %xmm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm18, (%rsi)
-; AVX512F-SLOW-NEXT: vmovdqa64 %ymm17, (%rdx)
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm17, (%rsi)
+; AVX512F-SLOW-NEXT: vmovdqa64 %ymm18, (%rdx)
; AVX512F-SLOW-NEXT: vmovdqa64 %ymm19, (%rcx)
; AVX512F-SLOW-NEXT: vmovdqa64 %ymm20, (%r8)
; AVX512F-SLOW-NEXT: vmovdqa64 %ymm21, (%r9)
@@ -5644,226 +5612,227 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-FAST-LABEL: load_i8_stride8_vf32:
; AVX512F-FAST: # %bb.0:
; AVX512F-FAST-NEXT: vmovdqa64 (%rdi), %zmm18
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,u,u,u,u,0,4,8,12>
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <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,0,4,8,12>
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [0,2,2,3,0,2,4,6]
; AVX512F-FAST-NEXT: vmovdqa64 224(%rdi), %ymm16
; AVX512F-FAST-NEXT: vpermd %ymm16, %ymm0, %ymm4
-; AVX512F-FAST-NEXT: vpshufb %ymm15, %ymm4, %ymm1
+; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm4, %ymm1
; AVX512F-FAST-NEXT: vmovdqa64 192(%rdi), %ymm17
-; AVX512F-FAST-NEXT: vpermd %ymm17, %ymm0, %ymm10
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,20,24,28,u,u,u,u]
+; AVX512F-FAST-NEXT: vpermd %ymm17, %ymm0, %ymm6
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,20,24,28,u,u,u,u]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
-; AVX512F-FAST-NEXT: vmovdqa64 160(%rdi), %ymm19
-; AVX512F-FAST-NEXT: vpermd %ymm19, %ymm0, %ymm9
+; AVX512F-FAST-NEXT: vmovdqa 160(%rdi), %ymm13
+; AVX512F-FAST-NEXT: vpermd %ymm13, %ymm0, %ymm9
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,20,24,28,u,u,u,u,u,u,u,u]
-; AVX512F-FAST-NEXT: vmovdqa64 128(%rdi), %ymm20
-; AVX512F-FAST-NEXT: vpermd %ymm20, %ymm0, %ymm11
+; AVX512F-FAST-NEXT: vmovdqa 128(%rdi), %ymm15
+; AVX512F-FAST-NEXT: vpermd %ymm15, %ymm0, %ymm11
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,20,24,28,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,u,u,u,u,0,8,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa 112(%rdi), %xmm12
-; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm12, %xmm5
-; AVX512F-FAST-NEXT: vmovdqa 96(%rdi), %xmm2
-; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm2, %xmm3
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,u,0,8,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa 80(%rdi), %xmm3
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm3, %xmm8
+; AVX512F-FAST-NEXT: vmovdqa 112(%rdi), %xmm1
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm1, %xmm5
+; AVX512F-FAST-NEXT: vmovdqa %xmm1, %xmm2
+; AVX512F-FAST-NEXT: vmovdqa 96(%rdi), %xmm1
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm1, %xmm3
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,u,u,0,8,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa 80(%rdi), %xmm12
+; AVX512F-FAST-NEXT: vpshufb %xmm8, %xmm12, %xmm10
; AVX512F-FAST-NEXT: vmovdqa 64(%rdi), %xmm5
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm5, %xmm7
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3]
-; AVX512F-FAST-NEXT: vpmovqb %zmm18, %xmm7
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm0, %ymm21
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <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,1,5,9,13>
-; AVX512F-FAST-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-FAST-NEXT: vmovdqa 32(%rdi), %ymm6
-; AVX512F-FAST-NEXT: vpsrlw $8, %ymm0, %ymm0
-; AVX512F-FAST-NEXT: vpsrlw $8, %ymm6, %ymm6
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm0, %zmm0
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm7
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm23
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm10, %ymm8
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm7[7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm9, %ymm14
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm25
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm11, %ymm7
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm24
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5],ymm7[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,u,u,u,u,1,9,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm8, %xmm12, %xmm14
-; AVX512F-FAST-NEXT: vpshufb %xmm8, %xmm2, %xmm8
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm14[0],xmm8[1],xmm14[1],xmm8[2],xmm14[2],xmm8[3],xmm14[3]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm14 = <u,u,u,u,1,9,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm14, %xmm3, %xmm13
-; AVX512F-FAST-NEXT: vpshufb %xmm14, %xmm5, %xmm14
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1],xmm14[2],xmm13[2],xmm14[3],xmm13[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm13[0,1,2],xmm8[3]
-; AVX512F-FAST-NEXT: vpmovqb %zmm0, %xmm0
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm8[2,3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
+; AVX512F-FAST-NEXT: vpshufb %xmm8, %xmm5, %xmm8
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm10[0],xmm8[1],xmm10[1],xmm8[2],xmm10[2],xmm8[3],xmm10[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm8[0,1,2],xmm7[3]
+; AVX512F-FAST-NEXT: vpmovqb %zmm18, %xmm8
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT: vmovdqa64 %ymm0, %ymm22
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <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,2,6,10,14>
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm31
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm10, %ymm7
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm4[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,17,21,25,29]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm7 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,17,21,25,29,u,u,u,u]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5,6],ymm0[7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm9, %ymm7
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm27
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm11, %ymm8
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm28
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3,4],ymm7[5],ymm8[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5],ymm0[6,7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,u,u,u,1,9,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm2, %xmm8
+; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm1, %xmm7
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,u,u,1,9,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %xmm8, %xmm12, %xmm10
+; AVX512F-FAST-NEXT: vpshufb %xmm8, %xmm5, %xmm8
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm10[0],xmm8[1],xmm10[1],xmm8[2],xmm10[2],xmm8[3],xmm10[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm8[0,1,2],xmm7[3]
+; AVX512F-FAST-NEXT: vpsrlq $8, %zmm18, %zmm8
+; AVX512F-FAST-NEXT: vpmovqb %zmm8, %xmm8
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm0, %ymm20
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,2,6,10,14>
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm4, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm29
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,2,6,10,14,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm6, %ymm7
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm30
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5,6],ymm0[7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm9, %ymm7
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm28
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm11, %ymm8
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm29
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm9, %ymm7
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm31
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm11, %ymm8
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm26
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3,4],ymm7[5],ymm8[6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,u,u,u,2,10,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm12, %xmm8
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm2, %xmm7
+; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm2, %xmm8
+; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm1, %xmm7
+; AVX512F-FAST-NEXT: vmovdqa %xmm1, %xmm3
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,u,u,2,10,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm8, %xmm3, %xmm13
+; AVX512F-FAST-NEXT: vpshufb %xmm8, %xmm12, %xmm10
; AVX512F-FAST-NEXT: vpshufb %xmm8, %xmm5, %xmm8
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm13[0],xmm8[1],xmm13[1],xmm8[2],xmm13[2],xmm8[3],xmm13[3]
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm10[0],xmm8[1],xmm10[1],xmm8[2],xmm10[2],xmm8[3],xmm10[3]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm8[0,1,2],xmm7[3]
-; AVX512F-FAST-NEXT: vpsrld $16, %zmm18, %zmm8
+; AVX512F-FAST-NEXT: vpsrlq $16, %zmm18, %zmm8
; AVX512F-FAST-NEXT: vpmovqb %zmm8, %xmm8
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm7[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm4[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,19,23,27,31]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm0, %ymm21
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <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,3,7,11,15>
+; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm25
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm10, %ymm4
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm27
+; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm6, %ymm4
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm24
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u,u,u,u,u>
; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm9, %ymm4
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm30
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm11, %ymm7
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm26
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0,1,2,3,4],ymm4[5],ymm7[6,7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm23
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm8 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm8, %ymm11, %ymm6
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3,4],ymm4[5],ymm6[6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,u,u,u,u,3,11,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm4, %xmm12, %xmm7
-; AVX512F-FAST-NEXT: vpshufb %xmm4, %xmm2, %xmm4
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1],xmm4[2],xmm7[2],xmm4[3],xmm7[3]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,u,3,11,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm3, %xmm9
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm5, %xmm7
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1],xmm7[2],xmm9[2],xmm7[3],xmm9[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
-; AVX512F-FAST-NEXT: vpsrld $24, %zmm18, %zmm7
-; AVX512F-FAST-NEXT: vpmovqb %zmm7, %xmm7
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-FAST-NEXT: vpshufb %xmm4, %xmm2, %xmm6
+; AVX512F-FAST-NEXT: vpshufb %xmm4, %xmm3, %xmm4
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm6 = <u,u,u,u,3,11,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %xmm6, %xmm12, %xmm9
+; AVX512F-FAST-NEXT: vpshufb %xmm6, %xmm5, %xmm6
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm6[0,1,2],xmm4[3]
+; AVX512F-FAST-NEXT: vpsrlq $24, %zmm18, %zmm6
+; AVX512F-FAST-NEXT: vpmovqb %zmm6, %xmm6
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm0, %ymm19
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [1,3,2,3,1,3,5,7]
; AVX512F-FAST-NEXT: vpermd %ymm16, %ymm0, %ymm11
-; AVX512F-FAST-NEXT: vpshufb %ymm15, %ymm11, %ymm7
+; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm11, %ymm4
; AVX512F-FAST-NEXT: vpermd %ymm17, %ymm0, %ymm9
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm13 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,20,24,28,u,u,u,u]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm13[0,1,2,3,4,5,6],ymm7[7]
-; AVX512F-FAST-NEXT: vpermd %ymm19, %ymm0, %ymm13
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm15 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,20,24,28,u,u,u,u,u,u,u,u]
-; AVX512F-FAST-NEXT: vpermd %ymm20, %ymm0, %ymm1
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm10 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,20,24,28,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm15[5],ymm10[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm10 = <u,u,u,u,u,u,4,12,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm10, %xmm12, %xmm15
-; AVX512F-FAST-NEXT: vpshufb %xmm10, %xmm2, %xmm10
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm10[0],xmm15[0],xmm10[1],xmm15[1],xmm10[2],xmm15[2],xmm10[3],xmm15[3]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm14 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,20,24,28,u,u,u,u]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1,2,3,4,5,6],ymm4[7]
+; AVX512F-FAST-NEXT: vpermd %ymm13, %ymm0, %ymm13
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm14 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,20,24,28,u,u,u,u,u,u,u,u]
+; AVX512F-FAST-NEXT: vpermd %ymm15, %ymm0, %ymm1
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm15 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,20,24,28,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5],ymm15[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1,2,3,4,5],ymm4[6,7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm14 = <u,u,u,u,u,u,4,12,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa %xmm2, %xmm6
+; AVX512F-FAST-NEXT: vpshufb %xmm14, %xmm2, %xmm15
+; AVX512F-FAST-NEXT: vpshufb %xmm14, %xmm3, %xmm14
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1],xmm14[2],xmm15[2],xmm14[3],xmm15[3]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm15 = <u,u,u,u,4,12,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm3, %xmm0
+; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm12, %xmm0
; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm5, %xmm15
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm15[0],xmm0[0],xmm15[1],xmm0[1],xmm15[2],xmm0[2],xmm15[3],xmm0[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm10[3]
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} zmm10 = zmm18[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
-; AVX512F-FAST-NEXT: vpmovqb %zmm10, %xmm10
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1],xmm0[2,3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm15 = ymm0[0,1,2,3],ymm7[4,5,6,7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm23, %ymm0
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm11, %ymm0
-; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm9, %ymm6
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5,6],ymm0[7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm25, %ymm6
-; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm13, %ymm6
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm24, %ymm7
-; AVX512F-FAST-NEXT: vpshufb %ymm7, %ymm1, %ymm7
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5],ymm7[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm6 = <u,u,u,u,u,u,5,13,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm6, %xmm12, %xmm7
-; AVX512F-FAST-NEXT: vpshufb %xmm6, %xmm2, %xmm6
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,u,5,13,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm3, %xmm10
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm5, %xmm7
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm10[0],xmm7[1],xmm10[1],xmm7[2],xmm10[2],xmm7[3],xmm10[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3]
-; AVX512F-FAST-NEXT: vpsrlq $40, %zmm18, %zmm7
-; AVX512F-FAST-NEXT: vpmovqb %zmm7, %xmm7
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm31, %ymm0
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm14[3]
+; AVX512F-FAST-NEXT: vpsrlq $32, %zmm18, %zmm14
+; AVX512F-FAST-NEXT: vpmovqb %zmm14, %xmm14
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm14[0,1],xmm0[2,3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm15 = ymm0[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm0 = 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,17,21,25,29]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm4 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,17,21,25,29,u,u,u,u]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5,6],ymm0[7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm27, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm13, %ymm4
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm28, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm14
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1,2,3,4],ymm4[5],ymm14[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,u,u,u,u,5,13,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %xmm4, %xmm6, %xmm14
+; AVX512F-FAST-NEXT: vmovdqa %xmm3, %xmm7
+; AVX512F-FAST-NEXT: vpshufb %xmm4, %xmm3, %xmm4
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm14[0],xmm4[1],xmm14[1],xmm4[2],xmm14[2],xmm4[3],xmm14[3]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm14 = <u,u,u,u,5,13,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %xmm14, %xmm12, %xmm0
+; AVX512F-FAST-NEXT: vpshufb %xmm14, %xmm5, %xmm14
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm14[0],xmm0[0],xmm14[1],xmm0[1],xmm14[2],xmm0[2],xmm14[3],xmm0[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm4[3]
+; AVX512F-FAST-NEXT: vpsrlq $40, %zmm18, %zmm4
+; AVX512F-FAST-NEXT: vpmovqb %zmm4, %xmm4
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm4[0,1],xmm0[2,3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm29, %ymm0
; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm11, %ymm0
-; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm9, %ymm7
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5,6],ymm0[7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm28, %ymm7
-; AVX512F-FAST-NEXT: vpshufb %ymm7, %ymm13, %ymm7
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm29, %ymm10
-; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm1, %ymm10
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm10[0,1,2,3,4],ymm7[5],ymm10[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5],ymm0[6,7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,u,u,u,6,14,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm12, %xmm10
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm2, %xmm7
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm10[0],xmm7[1],xmm10[1],xmm7[2],xmm10[2],xmm7[3],xmm10[3]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm10 = <u,u,u,u,6,14,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm10, %xmm3, %xmm14
-; AVX512F-FAST-NEXT: vpshufb %xmm10, %xmm5, %xmm10
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm10[0],xmm14[0],xmm10[1],xmm14[1],xmm10[2],xmm14[2],xmm10[3],xmm14[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm10[0,1,2],xmm7[3]
-; AVX512F-FAST-NEXT: vpsrlq $48, %zmm18, %zmm10
-; AVX512F-FAST-NEXT: vpmovqb %zmm10, %xmm10
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm7 = 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,19,23,27,31]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm27, %ymm10
-; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm9, %ymm9
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm9[0,1,2,3,4,5,6],ymm7[7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm9
-; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm13, %ymm9
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm26, %ymm10
-; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm1, %ymm1
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm9[5],ymm1[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm7[6,7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,u,u,u,7,15,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm12, %xmm9
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm2, %xmm2
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,u,7,15,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm3, %xmm3
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm5, %xmm5
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm9, %ymm2
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5,6],ymm0[7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm31, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm13, %ymm2
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm26, %ymm4
+; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm14
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0,1,2,3,4],ymm2[5],ymm14[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,u,u,u,u,6,14,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm6, %xmm14
+; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm2
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm14[0],xmm2[1],xmm14[1],xmm2[2],xmm14[2],xmm2[3],xmm14[3]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm14 = <u,u,u,u,6,14,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %xmm14, %xmm12, %xmm0
+; AVX512F-FAST-NEXT: vpshufb %xmm14, %xmm5, %xmm14
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm14[0],xmm0[0],xmm14[1],xmm0[1],xmm14[2],xmm0[2],xmm14[3],xmm0[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm2[3]
+; AVX512F-FAST-NEXT: vpsrlq $48, %zmm18, %zmm2
+; AVX512F-FAST-NEXT: vpmovqb %zmm2, %xmm2
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm25, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm11, %ymm2
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm24, %ymm3
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm9, %ymm3
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm23, %ymm3
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm13, %ymm3
+; AVX512F-FAST-NEXT: vpshufb %ymm8, %ymm1, %ymm1
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5],ymm1[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,u,u,u,u,7,15,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm6, %xmm3
+; AVX512F-FAST-NEXT: vpshufb %xmm2, %xmm7, %xmm2
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,u,u,7,15,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm12, %xmm8
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm5, %xmm3
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3]
; AVX512F-FAST-NEXT: vpsrlq $56, %zmm18, %zmm3
; AVX512F-FAST-NEXT: vpmovqb %zmm3, %xmm3
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, (%rsi)
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, (%rdx)
-; AVX512F-FAST-NEXT: vmovdqa %ymm8, (%rcx)
-; AVX512F-FAST-NEXT: vmovdqa %ymm4, (%r8)
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, (%rsi)
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm20, (%rdx)
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, (%rcx)
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, (%r8)
; AVX512F-FAST-NEXT: vmovdqa %ymm15, (%r9)
; AVX512F-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-FAST-NEXT: vmovdqa %ymm6, (%rax)
+; AVX512F-FAST-NEXT: vmovdqa %ymm10, (%rax)
; AVX512F-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT: vmovdqa %ymm0, (%rax)
; AVX512F-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
@@ -5949,7 +5918,7 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpshufb %xmm14, %xmm8, %xmm14
; AVX512BW-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1,2],xmm0[3]
-; AVX512BW-SLOW-NEXT: vpsrlw $8, %zmm16, %zmm10
+; AVX512BW-SLOW-NEXT: vpsrlq $8, %zmm16, %zmm10
; AVX512BW-SLOW-NEXT: vpmovqb %zmm10, %xmm10
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1],xmm0[2,3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
@@ -5984,7 +5953,7 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpshufb %xmm10, %xmm8, %xmm10
; AVX512BW-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm10[0],xmm14[0],xmm10[1],xmm14[1],xmm10[2],xmm14[2],xmm10[3],xmm14[3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1,2],xmm0[3]
-; AVX512BW-SLOW-NEXT: vpsrld $16, %zmm16, %zmm10
+; AVX512BW-SLOW-NEXT: vpsrlq $16, %zmm16, %zmm10
; AVX512BW-SLOW-NEXT: vpmovqb %zmm10, %xmm10
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1],xmm0[2,3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
@@ -6019,7 +5988,7 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpshufb %xmm10, %xmm8, %xmm10
; AVX512BW-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm10[0],xmm14[0],xmm10[1],xmm14[1],xmm10[2],xmm14[2],xmm10[3],xmm14[3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1,2],xmm0[3]
-; AVX512BW-SLOW-NEXT: vpsrld $24, %zmm16, %zmm10
+; AVX512BW-SLOW-NEXT: vpsrlq $24, %zmm16, %zmm10
; AVX512BW-SLOW-NEXT: vpmovqb %zmm10, %xmm10
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1],xmm0[2,3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
@@ -6054,7 +6023,7 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpshufb %xmm10, %xmm8, %xmm10
; AVX512BW-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm10[0],xmm14[0],xmm10[1],xmm14[1],xmm10[2],xmm14[2],xmm10[3],xmm14[3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1,2],xmm0[3]
-; AVX512BW-SLOW-NEXT: vpshufd {{.*#+}} zmm10 = zmm16[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512BW-SLOW-NEXT: vpsrlq $32, %zmm16, %zmm10
; AVX512BW-SLOW-NEXT: vpmovqb %zmm10, %xmm10
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1],xmm0[2,3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
@@ -6236,7 +6205,7 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpshufb %xmm8, %xmm5, %xmm8
; AVX512BW-FAST-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm20[0],xmm8[1],xmm20[1],xmm8[2],xmm20[2],xmm8[3],xmm20[3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm8[0,1,2],xmm7[3]
-; AVX512BW-FAST-NEXT: vpsrlw $8, %zmm0, %zmm8
+; AVX512BW-FAST-NEXT: vpsrlq $8, %zmm0, %zmm8
; AVX512BW-FAST-NEXT: vpmovqb %zmm8, %xmm8
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -6261,7 +6230,7 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpshufb %xmm10, %xmm5, %xmm10
; AVX512BW-FAST-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm10[0],xmm24[0],xmm10[1],xmm24[1],xmm10[2],xmm24[2],xmm10[3],xmm24[3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm10[0,1,2],xmm8[3]
-; AVX512BW-FAST-NEXT: vpsrld $16, %zmm0, %zmm10
+; AVX512BW-FAST-NEXT: vpsrlq $16, %zmm0, %zmm10
; AVX512BW-FAST-NEXT: vpmovqb %zmm10, %xmm10
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm10[0,1],xmm8[2,3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
@@ -6285,7 +6254,7 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpshufb %xmm10, %xmm5, %xmm10
; AVX512BW-FAST-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm10[0],xmm28[0],xmm10[1],xmm28[1],xmm10[2],xmm28[2],xmm10[3],xmm28[3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm10[0,1,2],xmm8[3]
-; AVX512BW-FAST-NEXT: vpsrld $24, %zmm0, %zmm10
+; AVX512BW-FAST-NEXT: vpsrlq $24, %zmm0, %zmm10
; AVX512BW-FAST-NEXT: vpmovqb %zmm10, %xmm10
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm10[0,1],xmm8[2,3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3],ymm4[4,5,6,7]
@@ -6310,7 +6279,7 @@ define void @load_i8_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpshufb %xmm13, %xmm5, %xmm13
; AVX512BW-FAST-NEXT: vpunpcklwd {{.*#+}} xmm13 = xmm13[0],xmm14[0],xmm13[1],xmm14[1],xmm13[2],xmm14[2],xmm13[3],xmm14[3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm13[0,1,2],xmm12[3]
-; AVX512BW-FAST-NEXT: vpshufd {{.*#+}} zmm13 = zmm0[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512BW-FAST-NEXT: vpsrlq $32, %zmm0, %zmm13
; AVX512BW-FAST-NEXT: vpmovqb %zmm13, %xmm13
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm13[0,1],xmm12[2,3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm12[0,1,2,3],ymm8[4,5,6,7]
@@ -11616,27 +11585,23 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
;
; AVX512F-SLOW-LABEL: load_i8_stride8_vf64:
; AVX512F-SLOW: # %bb.0:
-; AVX512F-SLOW-NEXT: subq $552, %rsp # imm = 0x228
-; AVX512F-SLOW-NEXT: vmovdqa64 256(%rdi), %zmm8
-; AVX512F-SLOW-NEXT: vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-SLOW-NEXT: subq $456, %rsp # imm = 0x1C8
+; AVX512F-SLOW-NEXT: vmovdqa64 256(%rdi), %zmm27
; AVX512F-SLOW-NEXT: vmovdqa64 384(%rdi), %zmm0
; AVX512F-SLOW-NEXT: vpmovqb %zmm0, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa 496(%rdi), %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa 496(%rdi), %xmm15
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,u,u,0,8,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm3, %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa %xmm3, %xmm7
-; AVX512F-SLOW-NEXT: vmovdqa 480(%rdi), %xmm12
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm12, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm15, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa 480(%rdi), %xmm4
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa %xmm4, %xmm9
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm3
-; AVX512F-SLOW-NEXT: vmovdqa 464(%rdi), %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa 464(%rdi), %xmm13
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,0,8,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa %xmm5, %xmm11
-; AVX512F-SLOW-NEXT: vmovdqa 448(%rdi), %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa %xmm6, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm13, %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa 448(%rdi), %xmm14
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm14, %xmm5
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5,6],ymm3[7]
@@ -11645,461 +11610,444 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: vpmovqb %ymm4, %xmm4
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4],ymm2[5],ymm4[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX512F-SLOW-NEXT: vmovdqa 368(%rdi), %xmm4
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa %xmm4, %xmm15
-; AVX512F-SLOW-NEXT: vmovdqa 352(%rdi), %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm4
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; AVX512F-SLOW-NEXT: vmovdqa 336(%rdi), %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa %xmm4, (%rsp) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm4, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa 320(%rdi), %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm6, %xmm21
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm4, %xmm31
+; AVX512F-SLOW-NEXT: vmovdqa 352(%rdi), %xmm11
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm11, %xmm4
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX512F-SLOW-NEXT: vmovdqa 336(%rdi), %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm19
+; AVX512F-SLOW-NEXT: vmovdqa 320(%rdi), %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm5
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3]
-; AVX512F-SLOW-NEXT: vpmovqb %zmm8, %xmm4
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3]
-; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm2
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3]
+; AVX512F-SLOW-NEXT: vpmovqb %zmm27, %xmm4
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm3, %zmm0, %zmm3
; AVX512F-SLOW-NEXT: movb $-64, %al
; AVX512F-SLOW-NEXT: kmovw %eax, %k1
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm2 {%k1}
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm3 {%k1}
; AVX512F-SLOW-NEXT: vmovdqa 240(%rdi), %xmm4
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa %xmm4, %xmm13
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm4, %xmm22
; AVX512F-SLOW-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vmovdqa 224(%rdi), %xmm5
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa 224(%rdi), %xmm12
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm12, %xmm4
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
+; AVX512F-SLOW-NEXT: vmovdqa 208(%rdi), %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm4
; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm24
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vmovdqa 208(%rdi), %xmm6
-; AVX512F-SLOW-NEXT: vmovdqa 256(%rdi), %ymm4
-; AVX512F-SLOW-NEXT: vpsrlw $8, %ymm4, %ymm4
-; AVX512F-SLOW-NEXT: vmovdqa 288(%rdi), %ymm5
-; AVX512F-SLOW-NEXT: vpsrlw $8, %ymm5, %ymm5
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm4, %zmm20
-; AVX512F-SLOW-NEXT: vmovdqa (%rdi), %ymm4
-; AVX512F-SLOW-NEXT: vmovdqa 32(%rdi), %ymm5
-; AVX512F-SLOW-NEXT: vpsrlw $8, %ymm4, %ymm4
-; AVX512F-SLOW-NEXT: vpsrlw $8, %ymm5, %ymm5
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm5, %zmm4, %zmm19
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm6, %xmm16
; AVX512F-SLOW-NEXT: vmovdqa 192(%rdi), %xmm6
; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm6, %xmm17
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm6, %xmm16
+; AVX512F-SLOW-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; AVX512F-SLOW-NEXT: vmovdqa64 128(%rdi), %zmm5
; AVX512F-SLOW-NEXT: vpmovqb %zmm5, %xmm5
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5,6],ymm3[7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm4
; AVX512F-SLOW-NEXT: vmovdqa 128(%rdi), %ymm5
; AVX512F-SLOW-NEXT: vpmovqb %ymm5, %xmm5
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5],ymm5[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-SLOW-NEXT: vmovdqa 112(%rdi), %xmm5
; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm29
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm20
; AVX512F-SLOW-NEXT: vmovdqa 96(%rdi), %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm29
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
; AVX512F-SLOW-NEXT: vmovdqa 80(%rdi), %xmm5
; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm22
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm21
; AVX512F-SLOW-NEXT: vmovdqa 64(%rdi), %xmm5
; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm23
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm30
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 (%rdi), %zmm1
-; AVX512F-SLOW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vpmovqb %zmm1, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 (%rdi), %zmm4
+; AVX512F-SLOW-NEXT: vpmovqb %zmm4, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %zmm4, %zmm18
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm2, %zmm0
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm3, %zmm0
; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,u,u,1,9,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa %xmm7, %xmm8
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm7, %xmm1
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm12, %xmm2
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,u,u,1,9,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm11, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm11, %xmm25
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,u,u,u,u,1,9,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm15, %xmm0
+; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm9, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm9, %xmm26
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm6 = <u,u,u,u,1,9,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm6, %xmm13, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm6, %xmm14, %xmm4
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
-; AVX512F-SLOW-NEXT: vmovdqa 416(%rdi), %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa 432(%rdi), %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm11 = <u,u,1,9,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm11, %xmm4, %xmm6
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm4, %xmm27
-; AVX512F-SLOW-NEXT: vpshufb %xmm11, %xmm3, %xmm7
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm3, %xmm31
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
-; AVX512F-SLOW-NEXT: vmovdqa 384(%rdi), %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa 400(%rdi), %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm4, %xmm9
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm4, %xmm30
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm3, %xmm14
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm3, %xmm28
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm14[0],xmm9[0],xmm14[1],xmm9[1],xmm14[2],xmm9[2],xmm14[3],xmm9[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5],ymm9[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1,2,3,4,5],ymm1[6,7]
-; AVX512F-SLOW-NEXT: vmovdqa %xmm15, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm15, %xmm9
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm3, %xmm10
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; AVX512F-SLOW-NEXT: vmovdqa (%rsp), %xmm3 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm9 = xmm10[0,1,2],xmm9[3]
-; AVX512F-SLOW-NEXT: vpmovqb %zmm20, %xmm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
-; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm9, %zmm0, %zmm14
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm14 {%k1}
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm13, %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm13
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm13, %xmm9
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm16, %xmm15
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm15, %xmm9
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm17, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm17, %xmm18
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0,1,2,3,4,5,6],ymm1[7]
-; AVX512F-SLOW-NEXT: vmovdqa 160(%rdi), %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa 176(%rdi), %xmm7
-; AVX512F-SLOW-NEXT: vpshufb %xmm11, %xmm7, %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm7, %xmm24
-; AVX512F-SLOW-NEXT: vpshufb %xmm11, %xmm3, %xmm11
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm3, %xmm20
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm11[0],xmm1[0],xmm11[1],xmm1[1],xmm11[2],xmm1[2],xmm11[3],xmm1[3]
-; AVX512F-SLOW-NEXT: vmovdqa 128(%rdi), %xmm7
-; AVX512F-SLOW-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vmovdqa 144(%rdi), %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm1, %xmm1
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm7, %xmm5
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm2[7]
+; AVX512F-SLOW-NEXT: vmovdqa 416(%rdi), %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa 432(%rdi), %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,1,9,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm1, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm25
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm0, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm0, %xmm23
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; AVX512F-SLOW-NEXT: vmovdqa 384(%rdi), %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vmovdqa 400(%rdi), %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <1,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm1, %xmm7
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm28
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm0, %xmm10
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm10[0],xmm7[0],xmm10[1],xmm7[1],xmm10[2],xmm7[2],xmm10[3],xmm7[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0,1,2,3,4],ymm5[5],ymm7[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm0
+; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm0, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm11, %xmm7
+; AVX512F-SLOW-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm9
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vpshufb %xmm6, %xmm9, %xmm7
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm6, %xmm0, %xmm10
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm10[0],xmm7[0],xmm10[1],xmm7[1],xmm10[2],xmm7[2],xmm10[3],xmm7[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm7[0,1,2],xmm5[3]
+; AVX512F-SLOW-NEXT: vpsrlq $8, %zmm27, %zmm7
+; AVX512F-SLOW-NEXT: vpmovqb %zmm7, %xmm7
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
+; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm5, %zmm0, %zmm19
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm19 {%k1}
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm0
+; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm0, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm12, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm12, %xmm22
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm12
+; AVX512F-SLOW-NEXT: vpshufb %xmm6, %xmm12, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm16, %xmm0
+; AVX512F-SLOW-NEXT: vpshufb %xmm6, %xmm0, %xmm7
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5,6],ymm3[7]
+; AVX512F-SLOW-NEXT: vmovdqa 160(%rdi), %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vmovdqa 176(%rdi), %xmm1
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm1, %xmm10
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm1, %xmm17
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm0, %xmm4
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm4[0],xmm10[0],xmm4[1],xmm10[1],xmm4[2],xmm10[2],xmm4[3],xmm10[3]
+; AVX512F-SLOW-NEXT: vmovdqa 128(%rdi), %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vmovdqa 144(%rdi), %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm16
+; AVX512F-SLOW-NEXT: vmovdqa %xmm5, (%rsp) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm4, %xmm2
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5],ymm1[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm3, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm0
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm7
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm7, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm9
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm2
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3]
-; AVX512F-SLOW-NEXT: vpmovqb %zmm19, %xmm2
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm14, %zmm0
-; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,u,u,2,10,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm8, %xmm1
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm8, %xmm26
-; AVX512F-SLOW-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm12, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm12, %xmm17
-; AVX512F-SLOW-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm7
+; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm7, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm2
+; AVX512F-SLOW-NEXT: vpshufb %xmm8, %xmm2, %xmm2
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,u,u,2,10,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm11
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm11, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm4, %xmm4
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,2,10,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm27, %xmm4
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm4, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm5
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm5, %xmm5
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm8
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm8, %xmm14
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm8
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm8, %xmm10
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm10[0],xmm14[0],xmm10[1],xmm14[1],xmm10[2],xmm14[2],xmm10[3],xmm14[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm10[0,1,2,3,4],ymm4[5],ymm10[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm6, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm6, %xmm22
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm6, %xmm10
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm10[0],xmm4[0],xmm10[1],xmm4[1],xmm10[2],xmm4[2],xmm10[3],xmm4[3]
-; AVX512F-SLOW-NEXT: vmovdqa (%rsp), %xmm8 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm8, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm12
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm10
; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm12, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm10[0,1,2],xmm4[3]
-; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm16 # 64-byte Reload
-; AVX512F-SLOW-NEXT: vpsrld $16, %zmm16, %zmm10
-; AVX512F-SLOW-NEXT: vpmovqb %zmm10, %xmm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm10[0,1],xmm4[2,3]
-; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm4, %zmm0, %zmm4
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm4 {%k1}
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm1, %xmm1
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm13, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm13, %xmm21
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm10[0],xmm1[0],xmm10[1],xmm1[1],xmm10[2],xmm1[2],xmm10[3],xmm1[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm15, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm15, %xmm19
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm18, %xmm13
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm13, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5,6],ymm1[7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm15
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm15, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm13
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm13, %xmm3
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm10, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm14, %xmm5
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm10[0],xmm5[1],xmm10[1],xmm5[2],xmm10[2],xmm5[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm3[5],ymm5[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm3, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm0
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm7, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm7, %xmm23
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm2
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3]
-; AVX512F-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm24 # 64-byte Reload
-; AVX512F-SLOW-NEXT: vpsrld $16, %zmm24, %zmm2
+; AVX512F-SLOW-NEXT: vpshufb %xmm6, %xmm10, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm6, %xmm3, %xmm3
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
+; AVX512F-SLOW-NEXT: vpsrlq $8, %zmm18, %zmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %zmm18, %zmm21
; AVX512F-SLOW-NEXT: vpmovqb %zmm2, %xmm2
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm4, %zmm0
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm19, %zmm0
; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,u,u,3,11,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,u,u,u,u,2,10,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm15, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm15, %xmm24
; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm0
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm17, %xmm2
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm2, %xmm2
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,3,11,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm11, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm17
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm0, %xmm2
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,2,10,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vmovdqa %xmm13, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm13, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm14, %xmm6
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm14, %xmm19
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,3,11,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm27, %xmm7
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm7, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm5
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm5, %xmm5
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm9
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm9, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm11
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm11, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm10[0,1,2,3,4],ymm4[5],ymm10[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm9
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm4
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm10
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm10[0],xmm4[0],xmm10[1],xmm4[1],xmm10[2],xmm4[2],xmm10[3],xmm4[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm8, %xmm10
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm12, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm10[0,1,2],xmm4[3]
-; AVX512F-SLOW-NEXT: vpsrld $24, %zmm16, %zmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %zmm16, %zmm20
-; AVX512F-SLOW-NEXT: vpmovqb %zmm10, %xmm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm10[0,1],xmm4[2,3]
-; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm4, %zmm0, %zmm4
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm4 {%k1}
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm6, %xmm16
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm10
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm10[0],xmm2[0],xmm10[1],xmm2[1],xmm10[2],xmm2[2],xmm10[3],xmm2[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm6, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm26
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm18, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm6, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1,2,3,4,5,6],ymm2[7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm15, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm15, %xmm19
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm13, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm13, %xmm22
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm12, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm12, %xmm18
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm3[0,1,2,3,4,5,6],ymm2[7]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,2,10,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm14
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm14, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm13
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm13, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1],xmm8[2],xmm2[2],xmm8[3],xmm2[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <2,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm0
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm0, %xmm0
; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm5, %xmm15, %xmm5
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm10[0],xmm5[1],xmm10[1],xmm5[2],xmm10[2],xmm5[3],xmm10[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm15, %xmm15
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm15[0],xmm0[0],xmm15[1],xmm0[1],xmm15[2],xmm0[2],xmm15[3],xmm0[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5],ymm0[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm6[6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm6, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm11, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm15
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
+; AVX512F-SLOW-NEXT: vpsrlq $16, %zmm27, %zmm8
+; AVX512F-SLOW-NEXT: vpmovqb %zmm8, %xmm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1],xmm6[2,3]
+; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm6, %zmm0, %zmm6
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm6 {%k1}
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm0, %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm11
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm11, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm12, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm12, %xmm20
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm12, %xmm15
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5,6],ymm0[7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm17, %xmm8
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm8, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm12, %xmm3
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm16, %xmm8
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm8, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm12, %xmm2
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm3[5],ymm5[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm13
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5],ymm2[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm7, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm7, %xmm18
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm7
; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm13, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm1
+; AVX512F-SLOW-NEXT: vpshufb %xmm4, %xmm7, %xmm3
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm10, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm10
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm10, %xmm1
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm12
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm12, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm0
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3]
+; AVX512F-SLOW-NEXT: vpsrlq $16, %zmm21, %zmm2
+; AVX512F-SLOW-NEXT: vpmovqb %zmm2, %xmm2
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm6, %zmm0
+; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,u,u,3,11,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm12
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm12, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm3, %xmm2
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,3,11,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm29
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm4
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm3[0,1,2,3,4,5,6],ymm2[7]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,3,11,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm14, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm13, %xmm6
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <3,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm13
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm13, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm14, %xmm15
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1,2,3,4],ymm6[5],ymm8[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm4[6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm6, %xmm6
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm8, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm8, %xmm8
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm15
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm9, %xmm22
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
+; AVX512F-SLOW-NEXT: vpsrlq $24, %zmm27, %zmm8
+; AVX512F-SLOW-NEXT: vpmovqb %zmm8, %xmm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1],xmm6[2,3]
+; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm6, %zmm0, %zmm6
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm6 {%k1}
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm4
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm11, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm8
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm8, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm15
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3,4,5,6],ymm4[7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm17, %xmm14
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm14, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm2
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vmovdqa (%rsp), %xmm8 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm8, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm9, %xmm3
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm18, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm3, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm7, %xmm0
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
-; AVX512F-SLOW-NEXT: vpsrld $24, %zmm24, %zmm1
-; AVX512F-SLOW-NEXT: vmovdqa64 %zmm24, %zmm29
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm3, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm10, %xmm1
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3]
+; AVX512F-SLOW-NEXT: vpsrlq $24, %zmm21, %zmm1
; AVX512F-SLOW-NEXT: vpmovqb %zmm1, %xmm1
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm4, %zmm0
+; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm6, %zmm0
; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,u,u,4,12,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm2, %xmm2
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,4,12,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm3, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm17, %xmm4
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,u,u,4,12,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm12, %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm2, %xmm2
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,4,12,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm3, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm4
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm3[0,1,2,3,4,5,6],ymm2[7]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,4,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm7, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm7
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm7, %xmm5
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
-; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm7
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm7, %xmm10
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm11, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm10[0,1,2,3,4],ymm5[5],ymm10[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa %xmm9, %xmm11
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,4,12,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm6
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <4,12,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm13, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm24
; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm10
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm10[0],xmm5[0],xmm10[1],xmm5[1],xmm10[2],xmm5[2],xmm10[3],xmm5[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm8, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm8, %xmm24
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm9, %xmm15
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1,2,3,4],ymm6[5],ymm8[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm4[6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm11
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm11, %xmm6
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm7, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm7, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm7
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm7, %xmm15
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm16
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
+; AVX512F-SLOW-NEXT: vpsrlq $32, %zmm27, %zmm8
+; AVX512F-SLOW-NEXT: vpmovqb %zmm8, %xmm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1],xmm6[2,3]
+; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm6, %zmm0, %zmm6
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm6 {%k1}
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm7, %xmm4
; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm8, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm10[0,1,2],xmm5[3]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} zmm10 = zmm20[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
-; AVX512F-SLOW-NEXT: vpmovqb %zmm10, %xmm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
-; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm5, %zmm0, %zmm5
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm5 {%k1}
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm16, %xmm8
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm8, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm9
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm25
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm10[0],xmm4[0],xmm10[1],xmm4[1],xmm10[2],xmm4[2],xmm10[3],xmm4[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm9
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm9, %xmm10
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm6, %xmm14
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm6, %xmm21
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm8, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm10
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm10, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm12, %xmm15
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm10[0,1,2,3,4,5,6],ymm4[7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm6, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm6, %xmm3
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm18, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm6, %xmm10
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm15, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm15, %xmm23
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3,4,5,6],ymm4[7]
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm14, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm13, %xmm2
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vmovdqa (%rsp), %xmm8 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm8, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm13, %xmm3
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5],ymm2[6,7]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm13, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm15, %xmm1
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm12, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm12, %xmm17
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm18, %xmm14
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm14, %xmm3
; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm0
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} zmm1 = zmm29[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm3, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm4, %xmm1
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3]
+; AVX512F-SLOW-NEXT: vpsrlq $32, %zmm21, %zmm1
+; AVX512F-SLOW-NEXT: vmovdqa64 %zmm21, %zmm22
; AVX512F-SLOW-NEXT: vpmovqb %zmm1, %xmm1
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm5, %zmm0
+; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm6, %zmm0
; AVX512F-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,u,u,5,13,u,u,u,u,u,u,u,u>
; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -12108,7 +12056,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm2, %xmm2
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,5,13,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm3
; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm3, %xmm3
; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm4, %xmm4
@@ -12117,82 +12065,79 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,5,13,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm27, %xmm9
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm6, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm27
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm13
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm13, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm6
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm25
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <5,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm7, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm13
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm13, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm10[0,1,2,3,4],ymm5[5],ymm10[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm11, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm6, %xmm10
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm10[0],xmm5[0],xmm10[1],xmm5[1],xmm10[2],xmm5[2],xmm10[3],xmm5[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm10[0,1,2],xmm5[3]
-; AVX512F-SLOW-NEXT: vpsrlq $40, %zmm20, %zmm10
-; AVX512F-SLOW-NEXT: vpmovqb %zmm10, %xmm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
-; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm5, %zmm0, %zmm5
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm5 {%k1}
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm8, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm16, %xmm31
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm6, %xmm10
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm10[0],xmm4[0],xmm10[1],xmm4[1],xmm10[2],xmm4[2],xmm10[3],xmm4[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm30
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm7
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm7, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm5, %xmm8
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm9, %xmm15
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1,2,3,4],ymm6[5],ymm8[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm4[6,7]
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm11, %xmm6
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm26
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm16, %xmm11
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm11, %xmm15
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
+; AVX512F-SLOW-NEXT: vpsrlq $40, %zmm27, %zmm8
+; AVX512F-SLOW-NEXT: vpmovqb %zmm8, %xmm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1],xmm6[2,3]
+; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm6, %zmm0, %zmm6
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm6 {%k1}
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm7, %xmm4
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm7, %xmm28
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm7, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm10, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm30
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm12, %xmm15
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm12, %xmm31
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm10[0,1,2,3,4,5,6],ymm4[7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm6, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm21
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm6, %xmm2
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm19
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm18, %xmm12
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm12, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm23, %xmm8
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm8, %xmm3
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3,4,5,6],ymm4[7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm17, %xmm11
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm11, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm2
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vmovdqa (%rsp), %xmm7 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm7, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm12, %xmm3
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm3, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm15, %xmm0
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm15, %xmm18
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm14, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm14, %xmm20
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm0
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm17, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm14, %xmm3
; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm4, %xmm1
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3]
-; AVX512F-SLOW-NEXT: vpsrlq $40, %zmm29, %zmm1
+; AVX512F-SLOW-NEXT: vpsrlq $40, %zmm21, %zmm1
; AVX512F-SLOW-NEXT: vpmovqb %zmm1, %xmm1
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm5, %zmm23
+; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm6, %zmm23
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,u,u,6,14,u,u,u,u,u,u,u,u>
; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm1, %xmm1
@@ -12200,8 +12145,9 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm2, %xmm2
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,6,14,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm3, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm4
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm4, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm21
; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm4, %xmm4
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
@@ -12209,80 +12155,78 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,6,14,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm9, %xmm28
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm27, %xmm5
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm5
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm13, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm13, %xmm16
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm6, %xmm6
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <6,14,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm9, %xmm10
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm13, %xmm14
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm13, %xmm26
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm10[0,1,2,3,4],ymm5[5],ymm10[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm11, %xmm5
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm11, %xmm16
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm13
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm13, %xmm8
; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm15, %xmm10
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm10[0],xmm5[0],xmm10[1],xmm5[1],xmm10[2],xmm5[2],xmm10[3],xmm5[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm9
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm11, %xmm14
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm10[0,1,2],xmm5[3]
-; AVX512F-SLOW-NEXT: vpsrlq $48, %zmm20, %zmm10
-; AVX512F-SLOW-NEXT: vpmovqb %zmm10, %xmm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
-; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm5, %zmm0, %zmm5
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm5 {%k1}
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm4
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm15, %xmm15
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1,2,3,4],ymm6[5],ymm8[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm4[6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm6, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm5, %xmm18
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm9, %xmm17
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm15
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
+; AVX512F-SLOW-NEXT: vpsrlq $48, %zmm27, %zmm8
+; AVX512F-SLOW-NEXT: vpmovqb %zmm8, %xmm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1],xmm6[2,3]
+; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm6, %zmm0, %zmm6
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm6 {%k1}
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm4
; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm4
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm13
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm13, %xmm10
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm10[0],xmm4[0],xmm10[1],xmm4[1],xmm10[2],xmm4[2],xmm10[3],xmm4[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm10
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm10, %xmm10
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm7, %xmm14
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm7, %xmm22
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm8
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm8, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm9
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm15
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm10[0,1,2,3,4,5,6],ymm4[7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm7
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm7, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm14
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm14, %xmm2
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm12, %xmm10
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm12, %xmm25
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm8, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm8, %xmm19
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3,4,5,6],ymm4[7]
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm11, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm11, %xmm19
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm2
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm10, %xmm24
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm7, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm7, %xmm29
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm12, %xmm3
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm12, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm18, %xmm8
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm8, %xmm0
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm20, %xmm10
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm10, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm11, %xmm0
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm6, %xmm17
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm10, %xmm1
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm14, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm7, %xmm1
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3]
-; AVX512F-SLOW-NEXT: vpsrlq $48, %zmm29, %zmm1
+; AVX512F-SLOW-NEXT: vpsrlq $48, %zmm22, %zmm1
; AVX512F-SLOW-NEXT: vpmovqb %zmm1, %xmm1
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm5, %zmm24
+; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm6, %zmm20
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,u,u,7,15,u,u,u,u,u,u,u,u>
; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm1, %xmm1
@@ -12290,7 +12234,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm2, %xmm2
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,7,15,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm3
; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm3, %xmm3
; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm4, %xmm4
@@ -12299,72 +12243,73 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,7,15,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm3
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm16, %xmm3
; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm27, %xmm5
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm5
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm6, %xmm6
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <7,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm6, %xmm6
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm7
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm7, %xmm7
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm13, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm9, %xmm9
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5],ymm6[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm16, %xmm5
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm5
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm15, %xmm6
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm11, %xmm7
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3]
-; AVX512F-SLOW-NEXT: vpsrlq $56, %zmm20, %zmm6
-; AVX512F-SLOW-NEXT: vpmovqb %zmm6, %xmm6
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
-; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm5, %zmm0, %zmm5
-; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm5 {%k1}
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm4
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1,2,3,4],ymm6[5],ymm8[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm4[6,7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm26, %xmm6
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm6, %xmm6
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm18, %xmm8
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm8, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm17, %xmm8
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm8, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm9, %xmm9
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
+; AVX512F-SLOW-NEXT: vpsrlq $56, %zmm27, %zmm8
+; AVX512F-SLOW-NEXT: vpmovqb %zmm8, %xmm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1],xmm6[2,3]
+; AVX512F-SLOW-NEXT: vinserti32x4 $2, %xmm6, %zmm0, %zmm6
+; AVX512F-SLOW-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm6 {%k1}
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm28, %xmm4
; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm4
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm13, %xmm6
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm6
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm22, %xmm7
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm7, %xmm7
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm8
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm30, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm8
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm31, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm9
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5,6],ymm4[7]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm21, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm6, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm14, %xmm2
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm25, %xmm6
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm6, %xmm6
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm7
-; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm7, %xmm3
-; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3]
+; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3,4,5,6],ymm4[7]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm19, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm5
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm24, %xmm8
+; AVX512F-SLOW-NEXT: vpshufb %xmm2, %xmm8, %xmm2
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
+; AVX512F-SLOW-NEXT: vmovdqa64 %xmm29, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm5, %xmm5
+; AVX512F-SLOW-NEXT: vpshufb %xmm3, %xmm12, %xmm3
+; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm12, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm8, %xmm0
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm10, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm0, %xmm11, %xmm0
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; AVX512F-SLOW-NEXT: vmovdqa64 %xmm17, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm3, %xmm3
-; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm10, %xmm1
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm14, %xmm3
+; AVX512F-SLOW-NEXT: vpshufb %xmm1, %xmm7, %xmm1
; AVX512F-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3]
-; AVX512F-SLOW-NEXT: vpsrlq $56, %zmm29, %zmm1
+; AVX512F-SLOW-NEXT: vpsrlq $56, %zmm22, %zmm1
; AVX512F-SLOW-NEXT: vpmovqb %zmm1, %xmm1
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm5, %zmm0
+; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm6, %zmm0
; AVX512F-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
; AVX512F-SLOW-NEXT: vmovaps %zmm1, (%rsi)
; AVX512F-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
@@ -12378,304 +12323,305 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT: vmovdqa64 %zmm23, (%rax)
; AVX512F-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-SLOW-NEXT: vmovdqa64 %zmm24, (%rax)
+; AVX512F-SLOW-NEXT: vmovdqa64 %zmm20, (%rax)
; AVX512F-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT: vmovdqa64 %zmm0, (%rax)
-; AVX512F-SLOW-NEXT: addq $552, %rsp # imm = 0x228
+; AVX512F-SLOW-NEXT: addq $456, %rsp # imm = 0x1C8
; AVX512F-SLOW-NEXT: vzeroupper
; AVX512F-SLOW-NEXT: retq
;
; AVX512F-FAST-LABEL: load_i8_stride8_vf64:
; AVX512F-FAST: # %bb.0:
-; AVX512F-FAST-NEXT: subq $440, %rsp # imm = 0x1B8
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <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,0,4,8,12>
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm8 = [0,2,2,3,0,2,4,6]
+; AVX512F-FAST-NEXT: subq $408, %rsp # imm = 0x198
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,0,4,8,12>
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [0,2,2,3,0,2,4,6]
; AVX512F-FAST-NEXT: vmovdqa 480(%rdi), %ymm0
; AVX512F-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vpermd %ymm0, %ymm8, %ymm5
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm5, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa %ymm1, %ymm9
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,0,4,8,12,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa 448(%rdi), %ymm1
-; AVX512F-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vpermd %ymm1, %ymm8, %ymm12
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm12, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa %ymm2, %ymm10
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa 416(%rdi), %ymm2
+; AVX512F-FAST-NEXT: vpermd %ymm0, %ymm1, %ymm11
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm11, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa %ymm2, %ymm9
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,0,4,8,12,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa 448(%rdi), %ymm2
; AVX512F-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vpermd %ymm2, %ymm8, %ymm3
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm3, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm16
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm18
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa 384(%rdi), %ymm3
+; AVX512F-FAST-NEXT: vpermd %ymm2, %ymm1, %ymm4
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm4, %ymm2
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm25
+; AVX512F-FAST-NEXT: vmovdqa %ymm3, %ymm10
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5,6],ymm0[7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa 416(%rdi), %ymm3
; AVX512F-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vpermd %ymm3, %ymm8, %ymm13
-; AVX512F-FAST-NEXT: vpshufb %ymm11, %ymm13, %ymm3
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,u,u,u,0,8,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa 368(%rdi), %xmm4
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm4, %xmm2
-; AVX512F-FAST-NEXT: vmovdqa 352(%rdi), %xmm14
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm14, %xmm3
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm15 = <u,u,u,u,0,8,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa 336(%rdi), %xmm1
-; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm1, %xmm3
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm1, %xmm30
-; AVX512F-FAST-NEXT: vmovdqa 320(%rdi), %xmm1
-; AVX512F-FAST-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm1, %xmm6
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3]
-; AVX512F-FAST-NEXT: vmovdqa64 256(%rdi), %zmm1
-; AVX512F-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vpmovqb %zmm1, %xmm3
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm17
+; AVX512F-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm4
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm4, %ymm3
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm24
+; AVX512F-FAST-NEXT: vmovdqa %ymm2, %ymm12
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 384(%rdi), %ymm31
+; AVX512F-FAST-NEXT: vpermd %ymm31, %ymm1, %ymm4
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm4, %ymm5
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm23
+; AVX512F-FAST-NEXT: vmovdqa %ymm2, %ymm15
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm3[5],ymm5[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,u,u,u,u,0,8,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa 368(%rdi), %xmm2
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm2, %xmm5
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm2, %xmm21
+; AVX512F-FAST-NEXT: vmovdqa 352(%rdi), %xmm13
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm13, %xmm6
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,u,u,0,8,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa 336(%rdi), %xmm2
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm2, %xmm7
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm2, %xmm27
+; AVX512F-FAST-NEXT: vmovdqa 320(%rdi), %xmm4
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm4, %xmm8
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3]
+; AVX512F-FAST-NEXT: vmovdqa64 256(%rdi), %zmm26
+; AVX512F-FAST-NEXT: vpmovqb %zmm26, %xmm7
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm6, %zmm0, %zmm16
; AVX512F-FAST-NEXT: movb $-64, %al
; AVX512F-FAST-NEXT: kmovw %eax, %k1
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm17 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa 224(%rdi), %ymm1
-; AVX512F-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa 256(%rdi), %ymm0
-; AVX512F-FAST-NEXT: vpsrlw $8, %ymm0, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa 288(%rdi), %ymm2
-; AVX512F-FAST-NEXT: vpsrlw $8, %ymm2, %ymm2
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm23
-; AVX512F-FAST-NEXT: vmovdqa (%rdi), %ymm0
-; AVX512F-FAST-NEXT: vmovdqa 32(%rdi), %ymm2
-; AVX512F-FAST-NEXT: vpsrlw $8, %ymm0, %ymm0
-; AVX512F-FAST-NEXT: vpsrlw $8, %ymm2, %ymm2
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm24
-; AVX512F-FAST-NEXT: vpermd %ymm1, %ymm8, %ymm0
-; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm0, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm0, %ymm22
-; AVX512F-FAST-NEXT: vmovdqa64 192(%rdi), %ymm29
-; AVX512F-FAST-NEXT: vpermd %ymm29, %ymm8, %ymm0
-; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm0, %ymm6
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm0, %ymm25
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm6[0,1,2,3,4,5,6],ymm2[7]
-; AVX512F-FAST-NEXT: vmovdqa64 160(%rdi), %ymm28
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm16 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa 224(%rdi), %ymm0
+; AVX512F-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-FAST-NEXT: vpermd %ymm0, %ymm1, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm2, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm22
+; AVX512F-FAST-NEXT: vmovdqa 192(%rdi), %ymm2
+; AVX512F-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-FAST-NEXT: vpermd %ymm2, %ymm1, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm2, %ymm6
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm20
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm0[7]
+; AVX512F-FAST-NEXT: vmovdqa 160(%rdi), %ymm0
+; AVX512F-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT: vmovdqa 128(%rdi), %ymm2
; AVX512F-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-FAST-NEXT: vpermd %ymm28, %ymm8, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm0
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm1, %ymm10
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm21
-; AVX512F-FAST-NEXT: vpermd %ymm2, %ymm8, %ymm0
-; AVX512F-FAST-NEXT: vpshufb %ymm11, %ymm0, %ymm8
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm0, %ymm26
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5],ymm8[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3,4,5],ymm9[6,7]
+; AVX512F-FAST-NEXT: vpermd %ymm0, %ymm1, %ymm14
+; AVX512F-FAST-NEXT: vpshufb %ymm12, %ymm14, %ymm7
+; AVX512F-FAST-NEXT: vpermd %ymm2, %ymm1, %ymm0
+; AVX512F-FAST-NEXT: vpshufb %ymm15, %ymm0, %ymm9
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm0, %ymm19
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm9[0,1,2,3,4],ymm7[5],ymm9[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm6[6,7]
; AVX512F-FAST-NEXT: vmovdqa 112(%rdi), %xmm0
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm0, %xmm8
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm0, %xmm27
-; AVX512F-FAST-NEXT: vmovdqa 96(%rdi), %xmm10
-; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm10, %xmm7
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
-; AVX512F-FAST-NEXT: vmovdqa 80(%rdi), %xmm9
-; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm9, %xmm0
-; AVX512F-FAST-NEXT: vmovdqa 64(%rdi), %xmm3
-; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm3, %xmm15
-; AVX512F-FAST-NEXT: vmovdqa %xmm3, %xmm8
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm15[0],xmm0[0],xmm15[1],xmm0[1],xmm15[2],xmm0[2],xmm15[3],xmm0[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
-; AVX512F-FAST-NEXT: vmovdqa64 (%rdi), %zmm31
-; AVX512F-FAST-NEXT: vpmovqb %zmm31, %xmm1
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm17, %zmm0
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm0, %xmm7
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm0, %xmm17
+; AVX512F-FAST-NEXT: vmovdqa 96(%rdi), %xmm0
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm0, %xmm3
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm0, %xmm29
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1],xmm3[2],xmm7[2],xmm3[3],xmm7[3]
+; AVX512F-FAST-NEXT: vmovdqa 80(%rdi), %xmm2
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm2, %xmm0
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm2, %xmm18
+; AVX512F-FAST-NEXT: vmovdqa 64(%rdi), %xmm15
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm15, %xmm5
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
+; AVX512F-FAST-NEXT: vmovdqa64 (%rdi), %zmm2
+; AVX512F-FAST-NEXT: vpmovqb %zmm2, %xmm3
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm2, %zmm28
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm16, %zmm0
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <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,1,5,9,13>
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm5, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm19
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,1,5,9,13>
+; AVX512F-FAST-NEXT: vpshufb %ymm7, %ymm11, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm11, %ymm16
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,1,5,9,13,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm12, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm20
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm25, %ymm10
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm10, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm30
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm16, %ymm7
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm7, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm17
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm13, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm18
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5],ymm2[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm24, %ymm11
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm11, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm24
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm23, %ymm8
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm8, %ymm3
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm25
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm1[5],ymm3[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,u,u,1,9,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm4, %xmm2
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm14, %xmm15
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm14, %xmm16
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm15[0],xmm2[0],xmm15[1],xmm2[1],xmm15[2],xmm2[2],xmm15[3],xmm2[3]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm15 = <u,u,u,u,1,9,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm30, %xmm6
-; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm6, %xmm0
-; AVX512F-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm11, %xmm14
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm14[0],xmm0[0],xmm14[1],xmm0[1],xmm14[2],xmm0[2],xmm14[3],xmm0[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm2[3]
-; AVX512F-FAST-NEXT: vpmovqb %zmm23, %xmm2
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm21, %xmm12
+; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm12, %xmm3
+; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm13, %xmm5
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm13, %xmm21
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,u,u,1,9,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm27, %xmm13
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm13, %xmm0
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm4, %xmm6
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm4, %xmm23
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm6[0],xmm0[0],xmm6[1],xmm0[1],xmm6[2],xmm0[2],xmm6[3],xmm0[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
+; AVX512F-FAST-NEXT: vpsrlq $8, %zmm26, %zmm3
+; AVX512F-FAST-NEXT: vpmovqb %zmm3, %xmm3
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm0, %zmm0, %zmm0
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, %ymm3
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm20, %ymm3
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm25, %ymm14
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm14, %ymm3
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, %ymm9
+; AVX512F-FAST-NEXT: vpshufb %ymm7, %ymm9, %ymm2
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm20, %ymm4
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm3
+; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm4, %ymm3
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm17, %ymm3
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, %ymm14
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm24, %ymm3
; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm14, %ymm3
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm14
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm26, %ymm11
-; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm11, %ymm14
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0,1,2,3,4],ymm3[5],ymm14[6,7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm14, %ymm20
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, %ymm7
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm25, %ymm6
+; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm7, %ymm6
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3,4],ymm3[5],ymm6[6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm27, %xmm11
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm11, %xmm3
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm10, %xmm1
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm17, %xmm6
+; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm6, %xmm3
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm17, %xmm22
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm29, %xmm6
+; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm6, %xmm1
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm29, %xmm17
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
-; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm9, %xmm3
-; AVX512F-FAST-NEXT: vpshufb %xmm15, %xmm8, %xmm14
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm14[0],xmm3[0],xmm14[1],xmm3[1],xmm14[2],xmm3[2],xmm14[3],xmm3[3]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm18, %xmm14
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm14, %xmm3
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm15, %xmm5
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
-; AVX512F-FAST-NEXT: vpmovqb %zmm24, %xmm3
+; AVX512F-FAST-NEXT: vpsrlq $8, %zmm28, %zmm3
+; AVX512F-FAST-NEXT: vpmovqb %zmm3, %xmm3
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <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,2,6,10,14>
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm5, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm16, %ymm0
+; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm0, %ymm0
; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm18
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm5, %ymm27
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,2,6,10,14,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm12, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm12, %ymm19
-; AVX512F-FAST-NEXT: vmovdqa %ymm2, %ymm12
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm10, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm10, %ymm24
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm29
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm7, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm17
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm13, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm23
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm11, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm11, %ymm25
+; AVX512F-FAST-NEXT: vmovdqa %ymm2, %ymm11
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm8, %ymm2
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm8, %ymm27
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5],ymm2[6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,u,u,2,10,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm4, %xmm2
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm4, %xmm30
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm16, %xmm4
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm4, %xmm3
+; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm12, %xmm2
+; AVX512F-FAST-NEXT: vmovdqa %xmm12, %xmm8
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm21, %xmm12
+; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm12, %xmm3
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,u,u,2,10,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm6, %xmm14
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm6, %xmm24
-; AVX512F-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm5, %xmm15
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm14 = xmm15[0],xmm14[0],xmm15[1],xmm14[1],xmm15[2],xmm14[2],xmm15[3],xmm14[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm14[0,1,2],xmm2[3]
-; AVX512F-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
-; AVX512F-FAST-NEXT: vpsrld $16, %zmm20, %zmm14
-; AVX512F-FAST-NEXT: vpmovqb %zmm14, %xmm14
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm14[0,1],xmm2[2,3]
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm13, %xmm5
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm13, %xmm19
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm23, %xmm13
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm13, %xmm6
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
+; AVX512F-FAST-NEXT: vpsrlq $16, %zmm26, %zmm5
+; AVX512F-FAST-NEXT: vpmovqb %zmm5, %xmm5
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm2
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm2 {%k1}
; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, %ymm6
-; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm6, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm25, %ymm6
-; AVX512F-FAST-NEXT: vpshufb %ymm12, %ymm6, %ymm14
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5,6],ymm0[7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm17, %ymm6
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, %ymm12
-; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm12, %ymm14
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm23, %ymm6
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm26, %ymm12
-; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm12, %ymm15
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5],ymm15[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5],ymm0[6,7]
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm11, %xmm14
+; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm9, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm9, %ymm18
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm29, %ymm5
+; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm21
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5,6],ymm0[7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm20, %ymm4
+; AVX512F-FAST-NEXT: vpshufb %ymm11, %ymm4, %ymm5
+; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm7, %ymm6
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm7, %ymm29
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5],ymm6[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm22, %xmm7
+; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm7, %xmm5
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm17, %xmm10
; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm10, %xmm1
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1],xmm1[2],xmm14[2],xmm1[3],xmm14[3]
-; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm9, %xmm14
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm9, %xmm18
-; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm8, %xmm3
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm8, %xmm16
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm14[0],xmm3[1],xmm14[1],xmm3[2],xmm14[2],xmm3[3],xmm14[3]
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm14, %xmm5
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm14, %xmm20
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm15, %xmm3
+; AVX512F-FAST-NEXT: vmovdqa %xmm15, %xmm9
+; AVX512F-FAST-NEXT: vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
-; AVX512F-FAST-NEXT: vpsrld $16, %zmm31, %zmm3
+; AVX512F-FAST-NEXT: vpsrlq $16, %zmm28, %zmm3
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm28, %zmm23
; AVX512F-FAST-NEXT: vpmovqb %zmm3, %xmm3
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm2, %zmm0
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <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,3,7,11,15>
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm27, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm16, %ymm0
; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm0, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa %ymm1, %ymm9
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,3,7,11,15,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, %ymm1
-; AVX512F-FAST-NEXT: vpshufb %ymm8, %ymm1, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm16
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,3,7,11,15,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm24, %ymm1
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm17
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm7, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm13, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm19
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm25, %ymm1
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa %ymm2, %ymm14
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm27, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm11, %ymm2, %ymm2
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5],ymm2[6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,u,u,3,11,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm30, %xmm15
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm15, %xmm2
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm4, %xmm3
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm4, %xmm17
+; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm8, %xmm2
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm8, %xmm22
+; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm12, %xmm3
+; AVX512F-FAST-NEXT: vmovdqa %xmm12, %xmm15
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,u,u,3,11,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm24, %xmm14
-; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm14, %xmm12
-; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm5, %xmm13
-; AVX512F-FAST-NEXT: vmovdqa %xmm5, %xmm7
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm12[0,1,2],xmm2[3]
-; AVX512F-FAST-NEXT: vpsrld $24, %zmm20, %zmm12
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm20, %zmm24
-; AVX512F-FAST-NEXT: vpmovqb %zmm12, %xmm12
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm12[0,1],xmm2[2,3]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm19, %xmm12
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm12, %xmm5
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm13, %xmm6
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
+; AVX512F-FAST-NEXT: vpsrlq $24, %zmm26, %zmm5
+; AVX512F-FAST-NEXT: vpmovqb %zmm5, %xmm5
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm2
; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm2 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, %ymm0
-; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm0, %ymm0
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm25, %ymm4
-; AVX512F-FAST-NEXT: vpshufb %ymm8, %ymm4, %ymm5
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm16, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm5
+; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm5, %ymm0
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm17, %ymm5
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, %ymm6
+; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm6, %ymm5
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5,6],ymm0[7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, %ymm4
-; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm4, %ymm4
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm26, %ymm5
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, %ymm6
-; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm5, %ymm5
+; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm4, %ymm4
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm29, %ymm5
+; AVX512F-FAST-NEXT: vpshufb %ymm11, %ymm5, %ymm5
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5],ymm5[6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
-; AVX512F-FAST-NEXT: vmovdqa %xmm11, %xmm9
-; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm11, %xmm4
+; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm7, %xmm4
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm7, %xmm17
; AVX512F-FAST-NEXT: vpshufb %xmm1, %xmm10, %xmm1
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm10, %xmm22
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm10, %xmm21
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm18, %xmm5
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm20, %xmm5
; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm5, %xmm4
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm18, %xmm23
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm16, %xmm5
-; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm5, %xmm3
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm16, %xmm25
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm9, %xmm3
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
-; AVX512F-FAST-NEXT: vpsrld $24, %zmm31, %zmm3
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm31, %zmm27
+; AVX512F-FAST-NEXT: vpsrlq $24, %zmm28, %zmm3
; AVX512F-FAST-NEXT: vpmovqb %zmm3, %xmm3
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -12683,256 +12629,256 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [1,3,2,3,1,3,5,7]
; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,0,4,8,12>
-; AVX512F-FAST-NEXT: vpshufb %ymm8, %ymm2, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm16
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm12, %ymm2
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,0,4,8,12>
+; AVX512F-FAST-NEXT: vpshufb %ymm7, %ymm2, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa %ymm2, %ymm4
; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm3, %ymm2
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm11, %ymm4, %ymm6
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm11, %ymm19
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm26
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3,4],ymm2[5],ymm6[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm11 = <u,u,u,u,u,u,4,12,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm15, %xmm2
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm30, %xmm18
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm17, %xmm4
-; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm4, %xmm6
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm13 = <u,u,u,u,4,12,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm14, %xmm6
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm14, %xmm17
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm7, %xmm14
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm7, %xmm31
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm14[0],xmm6[0],xmm14[1],xmm6[1],xmm14[2],xmm6[2],xmm14[3],xmm6[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm6[0,1,2],xmm2[3]
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} zmm6 = zmm20[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
-; AVX512F-FAST-NEXT: vpmovqb %zmm6, %xmm6
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm6[0,1],xmm2[2,3]
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm14
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm14 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm3, %ymm2
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: vpshufb %ymm8, %ymm2, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm21
-; AVX512F-FAST-NEXT: vpermd %ymm29, %ymm0, %ymm2
-; AVX512F-FAST-NEXT: vpshufb %ymm5, %ymm2, %ymm15
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm20
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm15 = ymm15[0,1,2,3,4,5,6],ymm1[7]
-; AVX512F-FAST-NEXT: vpermd %ymm28, %ymm0, %ymm2
-; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm2, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa %ymm2, %ymm10
-; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, %ymm2
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm5
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm0, %ymm30
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4],ymm1[5],ymm5[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm15[6,7]
-; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm9, %xmm5
-; AVX512F-FAST-NEXT: vmovdqa %xmm9, %xmm7
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm22, %xmm8
-; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm8, %xmm11
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm23, %xmm0
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm0, %xmm11
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm25, %xmm0
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm0, %xmm13
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm13[0],xmm11[0],xmm13[1],xmm11[1],xmm13[2],xmm11[2],xmm13[3],xmm11[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm5 = xmm11[0,1,2],xmm5[3]
-; AVX512F-FAST-NEXT: vpshufd {{.*#+}} zmm11 = zmm27[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm2, %ymm6
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm28
+; AVX512F-FAST-NEXT: vpermd %ymm31, %ymm0, %ymm2
+; AVX512F-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm31 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm31, %ymm8
+; AVX512F-FAST-NEXT: vpshufb %ymm8, %ymm2, %ymm8
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1,2,3,4],ymm6[5],ymm8[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm6 = <u,u,u,u,u,u,4,12,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm22, %xmm2
+; AVX512F-FAST-NEXT: vpshufb %xmm6, %xmm2, %xmm8
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm22, %xmm18
+; AVX512F-FAST-NEXT: vpshufb %xmm6, %xmm15, %xmm10
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm15, %xmm24
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm10[0],xmm8[0],xmm10[1],xmm8[1],xmm10[2],xmm8[2],xmm10[3],xmm8[3]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm11 = <u,u,u,u,4,12,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm12, %xmm10
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm19, %xmm22
+; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm13, %xmm12
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm13, %xmm16
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm12[0],xmm10[0],xmm12[1],xmm10[1],xmm12[2],xmm10[2],xmm12[3],xmm10[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm10[0,1,2],xmm8[3]
+; AVX512F-FAST-NEXT: vpsrlq $32, %zmm26, %zmm10
+; AVX512F-FAST-NEXT: vpmovqb %zmm10, %xmm10
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm10[0,1],xmm8[2,3]
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm8, %zmm0, %zmm12
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm12 {%k1}
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: vpshufb %ymm7, %ymm2, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm29
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm2, %ymm13
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm27
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm1[7]
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm1, %ymm15
+; AVX512F-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm31, %ymm0
+; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm1, %ymm14
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm1, %ymm30
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5],ymm14[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm17, %xmm9
+; AVX512F-FAST-NEXT: vpshufb %xmm6, %xmm9, %xmm14
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm21, %xmm7
+; AVX512F-FAST-NEXT: vpshufb %xmm6, %xmm7, %xmm6
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm14[0],xmm6[1],xmm14[1],xmm6[2],xmm14[2],xmm6[3],xmm14[3]
+; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm5, %xmm14
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm5, %xmm20
+; AVX512F-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm0, %xmm11
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm11[0],xmm14[0],xmm11[1],xmm14[1],xmm11[2],xmm14[2],xmm11[3],xmm14[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm11[0,1,2],xmm6[3]
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm23, %zmm19
+; AVX512F-FAST-NEXT: vpsrlq $32, %zmm23, %zmm11
; AVX512F-FAST-NEXT: vpmovqb %zmm11, %xmm11
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm5 = xmm11[0,1],xmm5[2,3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
-; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm1, %zmm14, %zmm0
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm11[0,1],xmm6[2,3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm13[4,5,6,7]
+; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm6, %zmm12, %zmm0
; AVX512F-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,1,5,9,13>
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm16, %ymm0
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm29
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,1,5,9,13,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm12, %ymm5
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm2, %ymm22
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5,6],ymm1[7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm3, %ymm5
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm19
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm26, %ymm2
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm2, %ymm11
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm3, %ymm16
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm11[0,1,2,3,4],ymm5[5],ymm11[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,u,u,u,u,5,13,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm18, %xmm3
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm3, %xmm11
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm4, %xmm13
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm4, %xmm28
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm13[0],xmm11[0],xmm13[1],xmm11[1],xmm13[2],xmm11[2],xmm13[3],xmm11[3]
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm4, %ymm6
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm25
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm21 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, %ymm1
+; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm3, %ymm11
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm11[0,1,2,3,4,5,6],ymm6[7]
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm17 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm17, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm28, %ymm0
+; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm0, %ymm11
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm12
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm12[0,1,2,3,4],ymm11[5],ymm12[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm11[0,1,2,3,4,5],ymm6[6,7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm11 = <u,u,u,u,u,u,5,13,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm18, %xmm5
+; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm5, %xmm12
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm24, %xmm10
+; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm10, %xmm13
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm13 = <u,u,u,u,5,13,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm17, %xmm2
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm2, %xmm14
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm31, %xmm6
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm6, %xmm15
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm22, %xmm1
+; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm1, %xmm14
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm16, %xmm8
+; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm8, %xmm15
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm14 = xmm15[0],xmm14[0],xmm15[1],xmm14[1],xmm15[2],xmm14[2],xmm15[3],xmm14[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm14[0,1,2],xmm11[3]
-; AVX512F-FAST-NEXT: vpsrlq $40, %zmm24, %zmm14
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm14[0,1,2],xmm12[3]
+; AVX512F-FAST-NEXT: vpsrlq $40, %zmm26, %zmm14
; AVX512F-FAST-NEXT: vpmovqb %zmm14, %xmm14
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm14[0,1],xmm11[2,3]
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm11, %zmm0, %zmm11
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm11 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm29, %ymm4
-; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm20, %ymm4
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, %ymm14
-; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm4, %ymm14
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5,6],ymm1[7]
-; AVX512F-FAST-NEXT: vpshufb %ymm9, %ymm10, %ymm14
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm10, %ymm29
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm9
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm16, %ymm4
-; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm9, %ymm15
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm14[0,1],xmm12[2,3]
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm12, %zmm0, %zmm12
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm0, %zmm12 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm29, %ymm6
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm6, %ymm6
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, %ymm2
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm27, %ymm14
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm14, %ymm14
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm14[0,1,2,3,4,5,6],ymm6[7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm17, %ymm2
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm14, %ymm14
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm15
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5],ymm15[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm1[6,7]
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm7, %xmm14
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm7, %xmm30
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm8, %xmm5
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm8, %xmm31
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm14[0],xmm5[1],xmm14[1],xmm5[2],xmm14[2],xmm5[3],xmm14[3]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm23, %xmm4
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm4, %xmm14
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm23, %xmm16
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm25, %xmm4
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm4, %xmm13
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm25, %xmm17
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm14[0,1,2,3,4,5],ymm6[6,7]
+; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm9, %xmm14
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm9, %xmm21
+; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm7, %xmm11
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm7, %xmm23
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm11[0],xmm14[0],xmm11[1],xmm14[1],xmm11[2],xmm14[2],xmm11[3],xmm14[3]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm20, %xmm9
+; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm9, %xmm14
+; AVX512F-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm7, %xmm13
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm13 = xmm13[0],xmm14[0],xmm13[1],xmm14[1],xmm13[2],xmm14[2],xmm13[3],xmm14[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm5 = xmm13[0,1,2],xmm5[3]
-; AVX512F-FAST-NEXT: vpsrlq $40, %zmm27, %zmm13
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm13[0,1,2],xmm11[3]
+; AVX512F-FAST-NEXT: vpsrlq $40, %zmm19, %zmm13
; AVX512F-FAST-NEXT: vpmovqb %zmm13, %xmm13
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm5 = xmm13[0,1],xmm5[2,3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
-; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm1, %zmm11, %zmm22
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm13[0,1],xmm11[2,3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm11[0,1,2,3],ymm6[4,5,6,7]
+; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm6, %zmm12, %zmm31
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <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,2,6,10,14>
-; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm0, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm18
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm0, %ymm23
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,2,6,10,14,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm8, %ymm12, %ymm5
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5,6],ymm1[7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm19, %ymm4
-; AVX512F-FAST-NEXT: vpshufb %ymm7, %ymm4, %ymm5
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm26, %ymm0
-; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm0, %ymm11
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm11[0,1,2,3,4],ymm5[5],ymm11[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,u,u,u,u,6,14,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm3, %xmm11
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm3, %xmm19
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm28, %xmm3
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm3, %xmm13
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm13[0],xmm11[0],xmm13[1],xmm11[1],xmm13[2],xmm11[2],xmm13[3],xmm11[3]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm25, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm6
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm4, %ymm20
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,2,6,10,14,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm11
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm11[0,1,2,3,4,5,6],ymm6[7]
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm22 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, %ymm4
+; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm0, %ymm11
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm0, %ymm18
+; AVX512F-FAST-NEXT: vmovdqa64 {{.*#+}} ymm17 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,2,6,10,14,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm17, %ymm12
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpshufb %ymm12, %ymm4, %ymm12
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm12[0,1,2,3,4],ymm11[5],ymm12[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm11[0,1,2,3,4,5],ymm6[6,7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm11 = <u,u,u,u,u,u,6,14,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm5, %xmm12
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm5, %xmm16
+; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm10, %xmm13
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm24, %xmm28
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm13 = <u,u,u,u,6,14,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm2, %xmm14
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm2, %xmm26
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm6, %xmm15
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm6, %xmm25
+; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm1, %xmm14
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm1, %xmm24
+; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm8, %xmm15
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm14 = xmm15[0],xmm14[0],xmm15[1],xmm14[1],xmm15[2],xmm14[2],xmm15[3],xmm14[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm14[0,1,2],xmm11[3]
-; AVX512F-FAST-NEXT: vpsrlq $48, %zmm24, %zmm14
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm14[0,1,2],xmm12[3]
+; AVX512F-FAST-NEXT: vpsrlq $48, %zmm26, %zmm14
; AVX512F-FAST-NEXT: vpmovqb %zmm14, %xmm14
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm14[0,1],xmm11[2,3]
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm11, %zmm0, %zmm11
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm11 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm1
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm2, %ymm1
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm20, %ymm2
-; AVX512F-FAST-NEXT: vpshufb %ymm8, %ymm2, %ymm14
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5,6],ymm1[7]
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm29, %ymm6
-; AVX512F-FAST-NEXT: vpshufb %ymm7, %ymm6, %ymm14
-; AVX512F-FAST-NEXT: vpshufb %ymm10, %ymm9, %ymm15
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm9, %ymm20
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm14[0,1],xmm12[2,3]
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm12, %zmm0, %zmm12
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm0, %zmm12 {%k1}
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm29, %ymm10
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm20, %ymm0
+; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm10, %ymm6
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm27, %ymm1
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm14
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm14[0,1,2,3,4,5,6],ymm6[7]
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm22, %ymm2
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm14
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm17, %ymm2
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm0
+; AVX512F-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm15
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5],ymm15[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm1[6,7]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm30, %xmm8
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm8, %xmm14
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm31, %xmm3
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm3, %xmm5
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm14[0],xmm5[1],xmm14[1],xmm5[2],xmm14[2],xmm5[3],xmm14[3]
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm16, %xmm9
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm14[0,1,2,3,4,5],ymm6[6,7]
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm21, %xmm15
+; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm15, %xmm14
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm23, %xmm0
+; AVX512F-FAST-NEXT: vpshufb %xmm11, %xmm0, %xmm11
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm11[0],xmm14[0],xmm11[1],xmm14[1],xmm11[2],xmm14[2],xmm11[3],xmm14[3]
; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm9, %xmm14
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm17, %xmm10
-; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm10, %xmm13
+; AVX512F-FAST-NEXT: vpshufb %xmm13, %xmm7, %xmm13
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm13 = xmm13[0],xmm14[0],xmm13[1],xmm14[1],xmm13[2],xmm14[2],xmm13[3],xmm14[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm5 = xmm13[0,1,2],xmm5[3]
-; AVX512F-FAST-NEXT: vpsrlq $48, %zmm27, %zmm13
-; AVX512F-FAST-NEXT: vpmovqb %zmm13, %xmm13
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm5 = xmm13[0,1],xmm5[2,3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
-; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm1, %zmm11, %zmm16
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,3,7,11,15>
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm23, %ymm1
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm5
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,3,7,11,15,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm7, %ymm12, %ymm11
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm11[0,1,2,3,4,5,6],ymm5[7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm15, %ymm4, %ymm4
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm0, %ymm11
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm11[0,1,2,3,4],ymm4[5],ymm11[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,u,u,u,u,7,15,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm19, %xmm0
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm0, %xmm11
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm28, %xmm0
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm0, %xmm12
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1],xmm12[2],xmm11[2],xmm12[3],xmm11[3]
-; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm12 = <u,u,u,u,7,15,u,u,u,u,u,u,u,u,u,u>
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm26, %xmm0
-; AVX512F-FAST-NEXT: vpshufb %xmm12, %xmm0, %xmm13
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm25, %xmm0
-; AVX512F-FAST-NEXT: vpshufb %xmm12, %xmm0, %xmm14
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1],xmm14[2],xmm13[2],xmm14[3],xmm13[3]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm13[0,1,2],xmm11[3]
-; AVX512F-FAST-NEXT: vpsrlq $56, %zmm24, %zmm13
+; AVX512F-FAST-NEXT: vpsrlq $48, %zmm19, %zmm13
; AVX512F-FAST-NEXT: vpmovqb %zmm13, %xmm13
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm13[0,1],xmm11[2,3]
-; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm11, %zmm0, %zmm11
-; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm0, %zmm11 {%k1}
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm21, %ymm0
-; AVX512F-FAST-NEXT: vpshufb %ymm3, %ymm0, %ymm4
-; AVX512F-FAST-NEXT: vpshufb %ymm7, %ymm2, %ymm3
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
-; AVX512F-FAST-NEXT: vpshufb %ymm15, %ymm6, %ymm2
-; AVX512F-FAST-NEXT: vmovdqa64 %ymm20, %ymm0
-; AVX512F-FAST-NEXT: vpshufb %ymm1, %ymm0, %ymm0
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm8, %xmm2
-; AVX512F-FAST-NEXT: vmovdqa64 %xmm31, %xmm1
-; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm1, %xmm3
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm11[0,1,2,3],ymm6[4,5,6,7]
+; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm6, %zmm12, %zmm17
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,3,7,11,15>
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm25, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm2, %ymm5
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} 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,3,7,11,15,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm13, %ymm3, %ymm3
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm18, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm2, %ymm2
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,3,7,11,15,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm4, %ymm4
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4],ymm2[5],ymm4[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,u,u,u,u,7,15,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm16, %xmm4
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm4
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm28, %xmm5
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm5, %xmm5
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,u,u,7,15,u,u,u,u,u,u,u,u,u,u>
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm24, %xmm11
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm11, %xmm11
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm8, %xmm12
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1],xmm12[2],xmm11[2],xmm12[3],xmm11[3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm11[0,1,2],xmm4[3]
+; AVX512F-FAST-NEXT: vpsrlq $56, %zmm26, %zmm11
+; AVX512F-FAST-NEXT: vpmovqb %zmm11, %xmm11
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm11[0,1],xmm4[2,3]
+; AVX512F-FAST-NEXT: vinserti32x4 $2, %xmm4, %zmm0, %zmm4
+; AVX512F-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm4 {%k1}
+; AVX512F-FAST-NEXT: vpshufb %ymm0, %ymm10, %ymm2
+; AVX512F-FAST-NEXT: vpshufb %ymm13, %ymm1, %ymm8
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3,4,5,6],ymm2[7]
+; AVX512F-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX512F-FAST-NEXT: vpshufb %ymm14, %ymm0, %ymm1
+; AVX512F-FAST-NEXT: vmovdqa64 %ymm30, %ymm0
+; AVX512F-FAST-NEXT: vpshufb %ymm6, %ymm0, %ymm0
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm15, %xmm1
+; AVX512F-FAST-NEXT: vmovdqa64 %xmm23, %xmm2
+; AVX512F-FAST-NEXT: vpshufb %xmm3, %xmm2, %xmm2
+; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm9, %xmm2
+; AVX512F-FAST-NEXT: vpshufb %xmm5, %xmm7, %xmm3
; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; AVX512F-FAST-NEXT: vpshufb %xmm12, %xmm9, %xmm3
-; AVX512F-FAST-NEXT: vpshufb %xmm12, %xmm10, %xmm4
-; AVX512F-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3]
-; AVX512F-FAST-NEXT: vpsrlq $56, %zmm27, %zmm3
-; AVX512F-FAST-NEXT: vpmovqb %zmm3, %xmm3
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm11, %zmm0
-; AVX512F-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
-; AVX512F-FAST-NEXT: vmovaps %zmm2, (%rsi)
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
+; AVX512F-FAST-NEXT: vpsrlq $56, %zmm19, %zmm2
+; AVX512F-FAST-NEXT: vpmovqb %zmm2, %xmm2
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm4, %zmm0
+; AVX512F-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512F-FAST-NEXT: vmovaps %zmm1, (%rsi)
; AVX512F-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
; AVX512F-FAST-NEXT: vmovaps %zmm1, (%rdx)
; AVX512F-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
@@ -12942,12 +12888,12 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
; AVX512F-FAST-NEXT: vmovaps %zmm1, (%r9)
; AVX512F-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm22, (%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm31, (%rax)
; AVX512F-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-FAST-NEXT: vmovdqa64 %zmm16, (%rax)
+; AVX512F-FAST-NEXT: vmovdqa64 %zmm17, (%rax)
; AVX512F-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT: vmovdqa64 %zmm0, (%rax)
-; AVX512F-FAST-NEXT: addq $440, %rsp # imm = 0x1B8
+; AVX512F-FAST-NEXT: addq $408, %rsp # imm = 0x198
; AVX512F-FAST-NEXT: vzeroupper
; AVX512F-FAST-NEXT: retq
;
@@ -13100,7 +13046,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm24
; AVX512BW-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm24[0],xmm4[0],xmm24[1],xmm4[1],xmm24[2],xmm4[2],xmm24[3],xmm4[3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3]
-; AVX512BW-SLOW-NEXT: vpsrlw $8, %zmm6, %zmm4
+; AVX512BW-SLOW-NEXT: vpsrlq $8, %zmm6, %zmm4
; AVX512BW-SLOW-NEXT: vpmovqb %zmm4, %xmm4
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
; AVX512BW-SLOW-NEXT: vinserti32x4 $2, %xmm3, %zmm0, %zmm3
@@ -13146,7 +13092,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
; AVX512BW-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
-; AVX512BW-SLOW-NEXT: vpsrlw $8, %zmm15, %zmm2
+; AVX512BW-SLOW-NEXT: vpsrlq $8, %zmm15, %zmm2
; AVX512BW-SLOW-NEXT: vpmovqb %zmm2, %xmm2
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -13195,7 +13141,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm26[0],xmm5[0],xmm26[1],xmm5[1],xmm26[2],xmm5[2],xmm26[3],xmm5[3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
; AVX512BW-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
-; AVX512BW-SLOW-NEXT: vpsrld $16, %zmm13, %zmm5
+; AVX512BW-SLOW-NEXT: vpsrlq $16, %zmm13, %zmm5
; AVX512BW-SLOW-NEXT: vpmovqb %zmm5, %xmm5
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
; AVX512BW-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm2
@@ -13234,7 +13180,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
; AVX512BW-SLOW-NEXT: vmovdqa64 %zmm15, %zmm9
-; AVX512BW-SLOW-NEXT: vpsrld $16, %zmm15, %zmm3
+; AVX512BW-SLOW-NEXT: vpsrlq $16, %zmm15, %zmm3
; AVX512BW-SLOW-NEXT: vpmovqb %zmm3, %xmm3
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -13277,7 +13223,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpshufb %xmm3, %xmm27, %xmm30
; AVX512BW-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm30[0],xmm5[0],xmm30[1],xmm5[1],xmm30[2],xmm5[2],xmm30[3],xmm5[3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
-; AVX512BW-SLOW-NEXT: vpsrld $24, %zmm13, %zmm5
+; AVX512BW-SLOW-NEXT: vpsrlq $24, %zmm13, %zmm5
; AVX512BW-SLOW-NEXT: vpmovqb %zmm5, %xmm5
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
; AVX512BW-SLOW-NEXT: vinserti32x4 $2, %xmm2, %zmm0, %zmm2
@@ -13316,7 +13262,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpshufb %xmm3, %xmm28, %xmm3
; AVX512BW-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm3[0,1,2],xmm0[3]
-; AVX512BW-SLOW-NEXT: vpsrld $24, %zmm9, %zmm3
+; AVX512BW-SLOW-NEXT: vpsrlq $24, %zmm9, %zmm3
; AVX512BW-SLOW-NEXT: vpmovqb %zmm3, %xmm3
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -13358,7 +13304,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpshufb %xmm2, %xmm12, %xmm30
; AVX512BW-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm30[0],xmm5[0],xmm30[1],xmm5[1],xmm30[2],xmm5[2],xmm30[3],xmm5[3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm5[0,1,2],xmm3[3]
-; AVX512BW-SLOW-NEXT: vpshufd {{.*#+}} zmm5 = zmm13[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512BW-SLOW-NEXT: vpsrlq $32, %zmm13, %zmm5
; AVX512BW-SLOW-NEXT: vpmovqb %zmm5, %xmm5
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3]
; AVX512BW-SLOW-NEXT: vinserti32x4 $2, %xmm3, %zmm0, %zmm3
@@ -13392,8 +13338,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpshufb %xmm2, %xmm23, %xmm2
; AVX512BW-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3]
-; AVX512BW-SLOW-NEXT: vpshufd $245, {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Folded Reload
-; AVX512BW-SLOW-NEXT: # zmm2 = mem[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512BW-SLOW-NEXT: vpsrlq $32, {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Folded Reload
; AVX512BW-SLOW-NEXT: vpmovqb %zmm2, %xmm2
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -13744,7 +13689,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpshufb %xmm24, %xmm29, %xmm21
; AVX512BW-FAST-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm21[0],xmm26[0],xmm21[1],xmm26[1],xmm21[2],xmm26[2],xmm21[3],xmm26[3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm12[0,1,2],xmm15[3]
-; AVX512BW-FAST-NEXT: vpsrlw $8, %zmm4, %zmm15
+; AVX512BW-FAST-NEXT: vpsrlq $8, %zmm4, %zmm15
; AVX512BW-FAST-NEXT: vpmovqb %zmm15, %xmm15
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm15[0,1],xmm12[2,3]
; AVX512BW-FAST-NEXT: vinserti32x4 $2, %xmm12, %zmm0, %zmm12
@@ -13763,7 +13708,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpshufb %xmm24, %xmm9, %xmm15
; AVX512BW-FAST-NEXT: vpunpcklwd {{.*#+}} xmm13 = xmm15[0],xmm13[0],xmm15[1],xmm13[1],xmm15[2],xmm13[2],xmm15[3],xmm13[3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm13[0,1,2],xmm11[3]
-; AVX512BW-FAST-NEXT: vpsrlw $8, %zmm10, %zmm13
+; AVX512BW-FAST-NEXT: vpsrlq $8, %zmm10, %zmm13
; AVX512BW-FAST-NEXT: vpmovqb %zmm13, %xmm13
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm13[0,1],xmm11[2,3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5,6,7]
@@ -13793,7 +13738,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpshufb %xmm13, %xmm29, %xmm21
; AVX512BW-FAST-NEXT: vpunpcklwd {{.*#+}} xmm15 = xmm21[0],xmm15[0],xmm21[1],xmm15[1],xmm21[2],xmm15[2],xmm21[3],xmm15[3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm15[0,1,2],xmm12[3]
-; AVX512BW-FAST-NEXT: vpsrld $16, %zmm4, %zmm15
+; AVX512BW-FAST-NEXT: vpsrlq $16, %zmm4, %zmm15
; AVX512BW-FAST-NEXT: vpmovqb %zmm15, %xmm15
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm15[0,1],xmm12[2,3]
; AVX512BW-FAST-NEXT: vinserti32x4 $2, %xmm12, %zmm0, %zmm12
@@ -13812,7 +13757,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpshufb %xmm13, %xmm9, %xmm13
; AVX512BW-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm13[0],xmm11[0],xmm13[1],xmm11[1],xmm13[2],xmm11[2],xmm13[3],xmm11[3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm11[0,1,2],xmm8[3]
-; AVX512BW-FAST-NEXT: vpsrld $16, %zmm10, %zmm11
+; AVX512BW-FAST-NEXT: vpsrlq $16, %zmm10, %zmm11
; AVX512BW-FAST-NEXT: vpmovqb %zmm11, %xmm11
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm11[0,1],xmm8[2,3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -13840,7 +13785,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpshufb %xmm12, %xmm29, %xmm15
; AVX512BW-FAST-NEXT: vpunpcklwd {{.*#+}} xmm14 = xmm15[0],xmm14[0],xmm15[1],xmm14[1],xmm15[2],xmm14[2],xmm15[3],xmm14[3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm14[0,1,2],xmm11[3]
-; AVX512BW-FAST-NEXT: vpsrld $24, %zmm4, %zmm14
+; AVX512BW-FAST-NEXT: vpsrlq $24, %zmm4, %zmm14
; AVX512BW-FAST-NEXT: vmovdqa64 %zmm4, %zmm17
; AVX512BW-FAST-NEXT: vpmovqb %zmm14, %xmm14
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm14[0,1],xmm11[2,3]
@@ -13860,7 +13805,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpshufb %xmm12, %xmm9, %xmm3
; AVX512BW-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
-; AVX512BW-FAST-NEXT: vpsrld $24, %zmm10, %zmm2
+; AVX512BW-FAST-NEXT: vpsrlq $24, %zmm10, %zmm2
; AVX512BW-FAST-NEXT: vpmovqb %zmm2, %xmm2
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -13892,7 +13837,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpshufb %xmm21, %xmm29, %xmm28
; AVX512BW-FAST-NEXT: vpunpcklwd {{.*#+}} xmm15 = xmm28[0],xmm15[0],xmm28[1],xmm15[1],xmm28[2],xmm15[2],xmm28[3],xmm15[3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm15[0,1,2],xmm12[3]
-; AVX512BW-FAST-NEXT: vpshufd {{.*#+}} zmm15 = zmm17[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512BW-FAST-NEXT: vpsrlq $32, %zmm17, %zmm15
; AVX512BW-FAST-NEXT: vpmovqb %zmm15, %xmm15
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm15[0,1],xmm12[2,3]
; AVX512BW-FAST-NEXT: vinserti32x4 $2, %xmm12, %zmm0, %zmm12
@@ -13915,7 +13860,7 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpshufb %xmm21, %xmm9, %xmm13
; AVX512BW-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm13[0],xmm11[0],xmm13[1],xmm11[1],xmm13[2],xmm11[2],xmm13[3],xmm11[3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm11[0,1,2],xmm8[3]
-; AVX512BW-FAST-NEXT: vpshufd {{.*#+}} zmm11 = zmm10[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
+; AVX512BW-FAST-NEXT: vpsrlq $32, %zmm10, %zmm11
; AVX512BW-FAST-NEXT: vpmovqb %zmm11, %xmm11
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm11[0,1],xmm8[2,3]
; AVX512BW-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1,2,3],ymm3[4,5,6,7]
@@ -14091,12 +14036,14 @@ define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; AVX: {{.*}}
; AVX2: {{.*}}
+; AVX512BW: {{.*}}
; AVX512BW-ONLY-FAST: {{.*}}
; AVX512BW-ONLY-SLOW: {{.*}}
; AVX512DQ-FAST: {{.*}}
; AVX512DQ-SLOW: {{.*}}
; AVX512DQBW-FAST: {{.*}}
; AVX512DQBW-SLOW: {{.*}}
+; AVX512F: {{.*}}
; AVX512F-ONLY-FAST: {{.*}}
; AVX512F-ONLY-SLOW: {{.*}}
; FALLBACK0: {{.*}}
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll b/llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll
index 4a7d8c155fcc..ac65790fbf58 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll
@@ -465,13 +465,13 @@ define <32 x i16> @shuffle_v32i16_32_zz_33_zz_34_zz_35_zz_36_zz_37_zz_38_zz_39_z
define <8 x i16> @pr32967(<32 x i16> %v) {
; KNL-LABEL: pr32967:
; KNL: ## %bb.0:
-; KNL-NEXT: vpsrld $16, %zmm0, %zmm0
+; KNL-NEXT: vpsrlq $16, %zmm0, %zmm0
; KNL-NEXT: vpmovqw %zmm0, %xmm0
; KNL-NEXT: retq
;
; SKX-LABEL: pr32967:
; SKX: ## %bb.0:
-; SKX-NEXT: vpsrld $16, %zmm0, %zmm0
+; SKX-NEXT: vpsrlq $16, %zmm0, %zmm0
; SKX-NEXT: vpmovqw %zmm0, %xmm0
; SKX-NEXT: vzeroupper
; SKX-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/x86-interleaved-access.ll b/llvm/test/CodeGen/X86/x86-interleaved-access.ll
index 490bd211eed8..ef9fe2c4daf4 100644
--- a/llvm/test/CodeGen/X86/x86-interleaved-access.ll
+++ b/llvm/test/CodeGen/X86/x86-interleaved-access.ll
@@ -529,7 +529,7 @@ define <16 x i1> @interleaved_load_vf16_i8_stride4(ptr %ptr) nounwind {
; AVX512: # %bb.0:
; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT: vpmovdb %zmm0, %xmm1
-; AVX512-NEXT: vpsrlw $8, %zmm0, %zmm2
+; AVX512-NEXT: vpsrld $8, %zmm0, %zmm2
; AVX512-NEXT: vpmovdb %zmm2, %xmm2
; AVX512-NEXT: vpsrld $16, %zmm0, %zmm3
; AVX512-NEXT: vpmovdb %zmm3, %xmm3
More information about the llvm-commits
mailing list