[llvm] [X86] shouldReduceLoadWidth - don't split loads if we can freely reuse full width legal binop (PR #129695)
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Tue Apr 29 01:29:54 PDT 2025
https://github.com/RKSimon updated https://github.com/llvm/llvm-project/pull/129695
>From 878a9575f00cf9d9bdf1b2749d1bd03dec8cc2a6 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev at redking.me.uk>
Date: Tue, 4 Mar 2025 12:30:38 +0000
Subject: [PATCH] [X86] shouldReduceLoadWidth - don't split loads if ANY uses
are a extract+store or a full width legal binop
Currently shouldReduceLoadWidth is very relaxed about when loads can be split to avoid extractions from the original full width load - resulting in many cases where the number of memory operations notably increases, replacing the cost of a extract_subvector for additional loads.
This patch adjusts the 256/512-bit vector load splitting metric to not split if ANY use of the full width load can be used directly - either in an extract+store (previously ALL uses had to be extract+store to prevent splits) or is used by a legal binop (so unlikely to be split itself).
This required a number of fixes - shouldReduceLoadWidth now needs to peek through bitcasts UP the use-chain to find final users (limited to hasOneUse cases to reduce complexity). It also exposed an issue in isTargetCanonicalConstantNode which assumed that a load of vector constant data would always extract, which is no longer the case.
---
llvm/lib/Target/X86/X86ISelLowering.cpp | 52 +-
llvm/lib/Target/X86/X86ISelLowering.h | 7 +-
llvm/test/CodeGen/X86/oddsubvector.ll | 8 +-
llvm/test/CodeGen/X86/setcc-lowering.ll | 2 +-
.../vector-interleaved-store-i16-stride-3.ll | 264 +--
.../vector-interleaved-store-i16-stride-5.ll | 388 ++--
.../vector-interleaved-store-i16-stride-6.ll | 1740 ++++++++---------
.../vector-interleaved-store-i16-stride-7.ll | 1498 +++++++-------
.../vector-interleaved-store-i8-stride-5.ll | 766 ++++----
.../vector-interleaved-store-i8-stride-6.ll | 200 +-
.../vector-interleaved-store-i8-stride-7.ll | 4 +-
llvm/test/CodeGen/X86/vselect-pcmp.ll | 4 +-
12 files changed, 2480 insertions(+), 2453 deletions(-)
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 4a9121baba7db..f4b2a189d1fae 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -3263,6 +3263,12 @@ bool X86TargetLowering::shouldReduceLoadWidth(
std::optional<unsigned> ByteOffset) const {
assert(cast<LoadSDNode>(Load)->isSimple() && "illegal to narrow");
+ auto PeekThroughOneUserBitcasts = [](const SDNode *N) {
+ while (N->getOpcode() == ISD::BITCAST && N->hasOneUse())
+ N = *N->user_begin();
+ return N;
+ };
+
// "ELF Handling for Thread-Local Storage" specifies that R_X86_64_GOTTPOFF
// relocation target a movq or addq instruction: don't let the load shrink.
SDValue BasePtr = cast<LoadSDNode>(Load)->getBasePtr();
@@ -3272,24 +3278,46 @@ bool X86TargetLowering::shouldReduceLoadWidth(
// If this is an (1) AVX vector load with (2) multiple uses and (3) all of
// those uses are extracted directly into a store, then the extract + store
- // can be store-folded. Therefore, it's probably not worth splitting the load.
+ // can be store-folded, or (4) any use will be used by legal full width
+ // instruction. Then, it's probably not worth splitting the load.
EVT VT = Load->getValueType(0);
if ((VT.is256BitVector() || VT.is512BitVector()) &&
!SDValue(Load, 0).hasOneUse()) {
+ bool FullWidthUse = false;
+ bool AllExtractStores = true;
for (SDUse &Use : Load->uses()) {
// Skip uses of the chain value. Result 0 of the node is the load value.
if (Use.getResNo() != 0)
continue;
- SDNode *User = Use.getUser();
+ const SDNode *User = PeekThroughOneUserBitcasts(Use.getUser());
- // If this use is not an extract + store, it's probably worth splitting.
- if (User->getOpcode() != ISD::EXTRACT_SUBVECTOR || !User->hasOneUse() ||
- User->user_begin()->getOpcode() != ISD::STORE)
- return true;
+ // If this use is an extract + store, it's probably not worth splitting.
+ if (User->getOpcode() == ISD::EXTRACT_SUBVECTOR &&
+ all_of(User->uses(), [&](const SDUse &U) {
+ const SDNode *Inner = PeekThroughOneUserBitcasts(U.getUser());
+ return Inner->getOpcode() == ISD::STORE;
+ }))
+ continue;
+
+ AllExtractStores = false;
+
+ // If any use is a full width legal/target bin op, then assume its legal
+ // and won't split.
+ if (isBinOp(User->getOpcode()) &&
+ (isOperationLegal(User->getOpcode(), User->getValueType(0)) ||
+ User->getOpcode() > ISD::BUILTIN_OP_END))
+ FullWidthUse = true;
}
- // All non-chain uses are extract + store.
- return false;
+
+ if (AllExtractStores)
+ return false;
+
+ // If we have an user that uses the full vector width, then this use is
+ // only worth splitting if the offset isn't 0 (to avoid an
+ // EXTRACT_SUBVECTOR) or we're loading a scalar integer.
+ if (FullWidthUse)
+ return (ByteOffset.value_or(0) > 0) || NewVT.isScalarInteger();
}
return true;
@@ -59134,6 +59162,14 @@ static SDValue combineINSERT_SUBVECTOR(SDNode *N, SelectionDAG &DAG,
return Res;
}
+ // Match insertion of subvector load that perfectly aliases a base load.
+ if ((IdxVal % SubVecNumElts) == 0 && ISD::isNormalLoad(Vec.getNode()) &&
+ ISD::isNormalLoad(SubVec.getNode()) &&
+ DAG.areNonVolatileConsecutiveLoads(
+ cast<LoadSDNode>(SubVec), cast<LoadSDNode>(Vec),
+ SubVec.getValueSizeInBits() / 8, IdxVal / SubVecNumElts))
+ return Vec;
+
return SDValue();
}
diff --git a/llvm/lib/Target/X86/X86ISelLowering.h b/llvm/lib/Target/X86/X86ISelLowering.h
index 7926292fc5bcf..662552a972249 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.h
+++ b/llvm/lib/Target/X86/X86ISelLowering.h
@@ -1337,8 +1337,8 @@ namespace llvm {
unsigned Depth) const override;
bool isTargetCanonicalConstantNode(SDValue Op) const override {
- // Peek through bitcasts/extracts/inserts to see if we have a broadcast
- // vector from memory.
+ // Peek through bitcasts/extracts/inserts to see if we have a vector
+ // load/broadcast from memory.
while (Op.getOpcode() == ISD::BITCAST ||
Op.getOpcode() == ISD::EXTRACT_SUBVECTOR ||
(Op.getOpcode() == ISD::INSERT_SUBVECTOR &&
@@ -1346,6 +1346,9 @@ namespace llvm {
Op = Op.getOperand(Op.getOpcode() == ISD::INSERT_SUBVECTOR ? 1 : 0);
return Op.getOpcode() == X86ISD::VBROADCAST_LOAD ||
+ Op.getOpcode() == X86ISD::SUBV_BROADCAST_LOAD ||
+ (Op.getOpcode() == ISD::LOAD &&
+ getTargetConstantFromLoad(cast<LoadSDNode>(Op))) ||
TargetLowering::isTargetCanonicalConstantNode(Op);
}
diff --git a/llvm/test/CodeGen/X86/oddsubvector.ll b/llvm/test/CodeGen/X86/oddsubvector.ll
index 2f557679a1558..a1da40e7e7655 100644
--- a/llvm/test/CodeGen/X86/oddsubvector.ll
+++ b/llvm/test/CodeGen/X86/oddsubvector.ll
@@ -261,9 +261,9 @@ define void @PR42833() {
;
; AVX2-LABEL: PR42833:
; AVX2: # %bb.0:
-; AVX2-NEXT: movl b(%rip), %eax
; AVX2-NEXT: vmovdqu c+128(%rip), %ymm0
-; AVX2-NEXT: addl c+128(%rip), %eax
+; AVX2-NEXT: vmovd %xmm0, %eax
+; AVX2-NEXT: addl b(%rip), %eax
; AVX2-NEXT: vmovd %eax, %xmm1
; AVX2-NEXT: vpaddd %ymm1, %ymm0, %ymm2
; AVX2-NEXT: vpaddd %ymm0, %ymm0, %ymm3
@@ -284,10 +284,10 @@ define void @PR42833() {
;
; AVX512-LABEL: PR42833:
; AVX512: # %bb.0:
-; AVX512-NEXT: movl b(%rip), %eax
; AVX512-NEXT: vmovdqu c+128(%rip), %ymm0
; AVX512-NEXT: vmovdqu64 c+128(%rip), %zmm1
-; AVX512-NEXT: addl c+128(%rip), %eax
+; AVX512-NEXT: vmovd %xmm0, %eax
+; AVX512-NEXT: addl b(%rip), %eax
; AVX512-NEXT: vmovd %eax, %xmm2
; AVX512-NEXT: vpaddd %ymm2, %ymm0, %ymm2
; AVX512-NEXT: vpaddd %ymm0, %ymm0, %ymm0
diff --git a/llvm/test/CodeGen/X86/setcc-lowering.ll b/llvm/test/CodeGen/X86/setcc-lowering.ll
index cdf9e180345ed..5a72bf3e18bd8 100644
--- a/llvm/test/CodeGen/X86/setcc-lowering.ll
+++ b/llvm/test/CodeGen/X86/setcc-lowering.ll
@@ -11,7 +11,7 @@ define <8 x i16> @pr25080(<8 x i32> %a) nounwind {
; AVX1-LABEL: pr25080:
; AVX1: # %bb.0: # %entry
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}+16(%rip), %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpackssdw %xmm0, %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-3.ll
index 6d1ba933b9082..7dbff047e4f87 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-3.ll
@@ -1672,50 +1672,50 @@ define void @store_i16_stride3_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm4[0,1,2,3]
; AVX512-NEXT: vpmovsxbd {{.*#+}} zmm4 = [5,5,0,6,6,0,7,7,0,8,8,0,9,9,0,10]
; AVX512-NEXT: vpermd (%rdx), %zmm4, %zmm5
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (mem & (zmm5 ^ zmm3))
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm6 = [0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (zmm6 & (zmm5 ^ zmm3))
; AVX512-NEXT: vmovdqa 32(%rdi), %ymm3
-; AVX512-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128,20,21,128,128]
-; AVX512-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX512-NEXT: vmovdqa 32(%rsi), %ymm8
-; AVX512-NEXT: vmovdqa {{.*#+}} ymm9 = [10,11,0,1,128,128,12,13,2,3,128,128,14,15,4,5,128,128,16,17,28,29,128,128,18,19,18,19,128,128,20,21]
-; AVX512-NEXT: vpshufb %ymm9, %ymm8, %ymm8
-; AVX512-NEXT: vpor %ymm3, %ymm8, %ymm3
-; AVX512-NEXT: vmovdqa 48(%rdi), %xmm8
-; AVX512-NEXT: vmovdqa 48(%rsi), %xmm10
-; AVX512-NEXT: vpunpckhwd {{.*#+}} xmm11 = xmm10[4],xmm8[4],xmm10[5],xmm8[5],xmm10[6],xmm8[6],xmm10[7],xmm8[7]
-; AVX512-NEXT: vpshufb %xmm7, %xmm11, %xmm7
-; AVX512-NEXT: vprold $16, %xmm10, %xmm10
-; AVX512-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[1,1,2,2]
-; AVX512-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1],xmm10[2],xmm8[3,4],xmm10[5],xmm8[6,7]
-; AVX512-NEXT: vinserti128 $1, %xmm7, %ymm8, %ymm7
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm8 = [128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128,20,21,128,128]
+; AVX512-NEXT: vpshufb %ymm8, %ymm3, %ymm3
+; AVX512-NEXT: vmovdqa 32(%rsi), %ymm9
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm10 = [10,11,0,1,128,128,12,13,2,3,128,128,14,15,4,5,128,128,16,17,28,29,128,128,18,19,18,19,128,128,20,21]
+; AVX512-NEXT: vpshufb %ymm10, %ymm9, %ymm9
+; AVX512-NEXT: vpor %ymm3, %ymm9, %ymm3
+; AVX512-NEXT: vmovdqa 48(%rdi), %xmm9
+; AVX512-NEXT: vmovdqa 48(%rsi), %xmm11
+; AVX512-NEXT: vpunpckhwd {{.*#+}} xmm12 = xmm11[4],xmm9[4],xmm11[5],xmm9[5],xmm11[6],xmm9[6],xmm11[7],xmm9[7]
+; AVX512-NEXT: vpshufb %xmm7, %xmm12, %xmm7
+; AVX512-NEXT: vprold $16, %xmm11, %xmm11
+; AVX512-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[1,1,2,2]
+; AVX512-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm11[2],xmm9[3,4],xmm11[5],xmm9[6,7]
+; AVX512-NEXT: vinserti128 $1, %xmm7, %ymm9, %ymm7
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm7[0,1,2,3]
; AVX512-NEXT: vmovdqa (%rdx), %ymm7
-; AVX512-NEXT: vmovdqa 32(%rdx), %ymm8
-; AVX512-NEXT: vmovdqa {{.*#+}} ymm10 = [128,128,10,11,128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128]
-; AVX512-NEXT: vpshufb %ymm10, %ymm8, %ymm11
-; AVX512-NEXT: vpermd %ymm8, %ymm4, %ymm4
-; AVX512-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm4
-; AVX512-NEXT: vinserti64x4 $1, %ymm4, %zmm11, %zmm4
+; AVX512-NEXT: vmovdqa 32(%rdx), %ymm9
+; AVX512-NEXT: vpermd %ymm9, %ymm4, %ymm4
+; AVX512-NEXT: vpandn %ymm4, %ymm6, %ymm4
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,10,11,128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128]
+; AVX512-NEXT: vpshufb %ymm6, %ymm9, %ymm9
+; AVX512-NEXT: vinserti64x4 $1, %ymm4, %zmm9, %zmm4
; AVX512-NEXT: vpternlogq {{.*#+}} zmm4 = zmm4 | (zmm3 & mem)
; AVX512-NEXT: vmovdqa (%rdi), %ymm3
-; AVX512-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX512-NEXT: vmovdqa (%rsi), %ymm6
-; AVX512-NEXT: vpshufb %ymm9, %ymm6, %ymm6
-; AVX512-NEXT: vpor %ymm3, %ymm6, %ymm3
-; AVX512-NEXT: vprold $16, %xmm0, %xmm6
-; AVX512-NEXT: vpshufd {{.*#+}} xmm8 = xmm1[1,1,2,2]
-; AVX512-NEXT: vpblendw {{.*#+}} xmm6 = xmm8[0,1],xmm6[2],xmm8[3,4],xmm6[5],xmm8[6,7]
+; AVX512-NEXT: vpshufb %ymm8, %ymm3, %ymm3
+; AVX512-NEXT: vmovdqa (%rsi), %ymm8
+; AVX512-NEXT: vpshufb %ymm10, %ymm8, %ymm8
+; AVX512-NEXT: vpor %ymm3, %ymm8, %ymm3
+; AVX512-NEXT: vprold $16, %xmm0, %xmm8
+; AVX512-NEXT: vpshufd {{.*#+}} xmm9 = xmm1[1,1,2,2]
+; AVX512-NEXT: vpblendw {{.*#+}} xmm8 = xmm9[0,1],xmm8[2],xmm9[3,4],xmm8[5],xmm9[6,7]
; AVX512-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512-NEXT: vpshufb %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm0
+; AVX512-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm0
; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm3[0,1,2,3]
-; AVX512-NEXT: vpshufb %ymm10, %ymm7, %ymm1
+; AVX512-NEXT: vpshufb %ymm6, %ymm7, %ymm1
; AVX512-NEXT: vpmovsxbd {{.*#+}} ymm2 = [0,0,0,0,1,1,0,2]
; AVX512-NEXT: vpermd %ymm7, %ymm2, %ymm2
-; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512-NEXT: vpandn %ymm2, %ymm3, %ymm2
+; AVX512-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm2
; AVX512-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 | (zmm0 & zmm3)
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 | (zmm0 & mem)
; AVX512-NEXT: vmovdqa64 %zmm1, (%rcx)
; AVX512-NEXT: vmovdqa64 %zmm4, 128(%rcx)
; AVX512-NEXT: vmovdqa64 %zmm5, 64(%rcx)
@@ -1747,50 +1747,50 @@ define void @store_i16_stride3_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm4[0,1,2,3]
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm4 = [5,5,0,6,6,0,7,7,0,8,8,0,9,9,0,10]
; AVX512-FCP-NEXT: vpermd (%rdx), %zmm4, %zmm5
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (mem & (zmm5 ^ zmm3))
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm6 = [0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (zmm6 & (zmm5 ^ zmm3))
; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %ymm3
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128,20,21,128,128]
-; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX512-FCP-NEXT: vmovdqa 32(%rsi), %ymm8
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm9 = [10,11,0,1,128,128,12,13,2,3,128,128,14,15,4,5,128,128,16,17,28,29,128,128,18,19,18,19,128,128,20,21]
-; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm8, %ymm8
-; AVX512-FCP-NEXT: vpor %ymm3, %ymm8, %ymm3
-; AVX512-FCP-NEXT: vmovdqa 48(%rdi), %xmm8
-; AVX512-FCP-NEXT: vmovdqa 48(%rsi), %xmm10
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm11 = xmm10[4],xmm8[4],xmm10[5],xmm8[5],xmm10[6],xmm8[6],xmm10[7],xmm8[7]
-; AVX512-FCP-NEXT: vpshufb %xmm7, %xmm11, %xmm7
-; AVX512-FCP-NEXT: vprold $16, %xmm10, %xmm10
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[1,1,2,2]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1],xmm10[2],xmm8[3,4],xmm10[5],xmm8[6,7]
-; AVX512-FCP-NEXT: vinserti128 $1, %xmm7, %ymm8, %ymm7
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm8 = [128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128,20,21,128,128]
+; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm3, %ymm3
+; AVX512-FCP-NEXT: vmovdqa 32(%rsi), %ymm9
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm10 = [10,11,0,1,128,128,12,13,2,3,128,128,14,15,4,5,128,128,16,17,28,29,128,128,18,19,18,19,128,128,20,21]
+; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm9, %ymm9
+; AVX512-FCP-NEXT: vpor %ymm3, %ymm9, %ymm3
+; AVX512-FCP-NEXT: vmovdqa 48(%rdi), %xmm9
+; AVX512-FCP-NEXT: vmovdqa 48(%rsi), %xmm11
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm12 = xmm11[4],xmm9[4],xmm11[5],xmm9[5],xmm11[6],xmm9[6],xmm11[7],xmm9[7]
+; AVX512-FCP-NEXT: vpshufb %xmm7, %xmm12, %xmm7
+; AVX512-FCP-NEXT: vprold $16, %xmm11, %xmm11
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[1,1,2,2]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm11[2],xmm9[3,4],xmm11[5],xmm9[6,7]
+; AVX512-FCP-NEXT: vinserti128 $1, %xmm7, %ymm9, %ymm7
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm7[0,1,2,3]
; AVX512-FCP-NEXT: vmovdqa (%rdx), %ymm7
-; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %ymm8
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm10 = [128,128,10,11,128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128]
-; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm8, %ymm11
-; AVX512-FCP-NEXT: vpermd %ymm8, %ymm4, %ymm4
-; AVX512-FCP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm4
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm11, %zmm4
+; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %ymm9
+; AVX512-FCP-NEXT: vpermd %ymm9, %ymm4, %ymm4
+; AVX512-FCP-NEXT: vpandn %ymm4, %ymm6, %ymm4
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,10,11,128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128]
+; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm9, %ymm9
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm9, %zmm4
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm4 | (zmm3 & mem)
; AVX512-FCP-NEXT: vmovdqa (%rdi), %ymm3
-; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX512-FCP-NEXT: vmovdqa (%rsi), %ymm6
-; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm6, %ymm6
-; AVX512-FCP-NEXT: vpor %ymm3, %ymm6, %ymm3
-; AVX512-FCP-NEXT: vprold $16, %xmm0, %xmm6
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm8 = xmm1[1,1,2,2]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm6 = xmm8[0,1],xmm6[2],xmm8[3,4],xmm6[5],xmm8[6,7]
+; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm3, %ymm3
+; AVX512-FCP-NEXT: vmovdqa (%rsi), %ymm8
+; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm8, %ymm8
+; AVX512-FCP-NEXT: vpor %ymm3, %ymm8, %ymm3
+; AVX512-FCP-NEXT: vprold $16, %xmm0, %xmm8
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm9 = xmm1[1,1,2,2]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm8 = xmm9[0,1],xmm8[2],xmm9[3,4],xmm8[5],xmm9[6,7]
; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm0, %xmm0
-; AVX512-FCP-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm0
+; AVX512-FCP-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm0
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm3[0,1,2,3]
-; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm7, %ymm1
+; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm7, %ymm1
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm2 = [0,0,0,0,1,1,0,2]
; AVX512-FCP-NEXT: vpermd %ymm7, %ymm2, %ymm2
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512-FCP-NEXT: vpandn %ymm2, %ymm3, %ymm2
+; AVX512-FCP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm2
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 | (zmm0 & zmm3)
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 | (zmm0 & mem)
; AVX512-FCP-NEXT: vmovdqa64 %zmm1, (%rcx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm4, 128(%rcx)
; AVX512-FCP-NEXT: vmovdqa64 %zmm5, 64(%rcx)
@@ -1822,50 +1822,50 @@ define void @store_i16_stride3_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm4[0,1,2,3]
; AVX512DQ-NEXT: vpmovsxbd {{.*#+}} zmm4 = [5,5,0,6,6,0,7,7,0,8,8,0,9,9,0,10]
; AVX512DQ-NEXT: vpermd (%rdx), %zmm4, %zmm5
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (mem & (zmm5 ^ zmm3))
+; AVX512DQ-NEXT: vmovdqa64 {{.*#+}} zmm6 = [0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (zmm6 & (zmm5 ^ zmm3))
; AVX512DQ-NEXT: vmovdqa 32(%rdi), %ymm3
-; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128,20,21,128,128]
-; AVX512DQ-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX512DQ-NEXT: vmovdqa 32(%rsi), %ymm8
-; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm9 = [10,11,0,1,128,128,12,13,2,3,128,128,14,15,4,5,128,128,16,17,28,29,128,128,18,19,18,19,128,128,20,21]
-; AVX512DQ-NEXT: vpshufb %ymm9, %ymm8, %ymm8
-; AVX512DQ-NEXT: vpor %ymm3, %ymm8, %ymm3
-; AVX512DQ-NEXT: vmovdqa 48(%rdi), %xmm8
-; AVX512DQ-NEXT: vmovdqa 48(%rsi), %xmm10
-; AVX512DQ-NEXT: vpunpckhwd {{.*#+}} xmm11 = xmm10[4],xmm8[4],xmm10[5],xmm8[5],xmm10[6],xmm8[6],xmm10[7],xmm8[7]
-; AVX512DQ-NEXT: vpshufb %xmm7, %xmm11, %xmm7
-; AVX512DQ-NEXT: vprold $16, %xmm10, %xmm10
-; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[1,1,2,2]
-; AVX512DQ-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1],xmm10[2],xmm8[3,4],xmm10[5],xmm8[6,7]
-; AVX512DQ-NEXT: vinserti128 $1, %xmm7, %ymm8, %ymm7
+; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm8 = [128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128,20,21,128,128]
+; AVX512DQ-NEXT: vpshufb %ymm8, %ymm3, %ymm3
+; AVX512DQ-NEXT: vmovdqa 32(%rsi), %ymm9
+; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm10 = [10,11,0,1,128,128,12,13,2,3,128,128,14,15,4,5,128,128,16,17,28,29,128,128,18,19,18,19,128,128,20,21]
+; AVX512DQ-NEXT: vpshufb %ymm10, %ymm9, %ymm9
+; AVX512DQ-NEXT: vpor %ymm3, %ymm9, %ymm3
+; AVX512DQ-NEXT: vmovdqa 48(%rdi), %xmm9
+; AVX512DQ-NEXT: vmovdqa 48(%rsi), %xmm11
+; AVX512DQ-NEXT: vpunpckhwd {{.*#+}} xmm12 = xmm11[4],xmm9[4],xmm11[5],xmm9[5],xmm11[6],xmm9[6],xmm11[7],xmm9[7]
+; AVX512DQ-NEXT: vpshufb %xmm7, %xmm12, %xmm7
+; AVX512DQ-NEXT: vprold $16, %xmm11, %xmm11
+; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[1,1,2,2]
+; AVX512DQ-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm11[2],xmm9[3,4],xmm11[5],xmm9[6,7]
+; AVX512DQ-NEXT: vinserti128 $1, %xmm7, %ymm9, %ymm7
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm7[0,1,2,3]
; AVX512DQ-NEXT: vmovdqa (%rdx), %ymm7
-; AVX512DQ-NEXT: vmovdqa 32(%rdx), %ymm8
-; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm10 = [128,128,10,11,128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128]
-; AVX512DQ-NEXT: vpshufb %ymm10, %ymm8, %ymm11
-; AVX512DQ-NEXT: vpermd %ymm8, %ymm4, %ymm4
-; AVX512DQ-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm4
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm4, %zmm11, %zmm4
+; AVX512DQ-NEXT: vmovdqa 32(%rdx), %ymm9
+; AVX512DQ-NEXT: vpermd %ymm9, %ymm4, %ymm4
+; AVX512DQ-NEXT: vpandn %ymm4, %ymm6, %ymm4
+; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,10,11,128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128]
+; AVX512DQ-NEXT: vpshufb %ymm6, %ymm9, %ymm9
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm4, %zmm9, %zmm4
; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm4 = zmm4 | (zmm3 & mem)
; AVX512DQ-NEXT: vmovdqa (%rdi), %ymm3
-; AVX512DQ-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX512DQ-NEXT: vmovdqa (%rsi), %ymm6
-; AVX512DQ-NEXT: vpshufb %ymm9, %ymm6, %ymm6
-; AVX512DQ-NEXT: vpor %ymm3, %ymm6, %ymm3
-; AVX512DQ-NEXT: vprold $16, %xmm0, %xmm6
-; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm8 = xmm1[1,1,2,2]
-; AVX512DQ-NEXT: vpblendw {{.*#+}} xmm6 = xmm8[0,1],xmm6[2],xmm8[3,4],xmm6[5],xmm8[6,7]
+; AVX512DQ-NEXT: vpshufb %ymm8, %ymm3, %ymm3
+; AVX512DQ-NEXT: vmovdqa (%rsi), %ymm8
+; AVX512DQ-NEXT: vpshufb %ymm10, %ymm8, %ymm8
+; AVX512DQ-NEXT: vpor %ymm3, %ymm8, %ymm3
+; AVX512DQ-NEXT: vprold $16, %xmm0, %xmm8
+; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm9 = xmm1[1,1,2,2]
+; AVX512DQ-NEXT: vpblendw {{.*#+}} xmm8 = xmm9[0,1],xmm8[2],xmm9[3,4],xmm8[5],xmm9[6,7]
; AVX512DQ-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512DQ-NEXT: vpshufb %xmm2, %xmm0, %xmm0
-; AVX512DQ-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm0
+; AVX512DQ-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm0
; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm3[0,1,2,3]
-; AVX512DQ-NEXT: vpshufb %ymm10, %ymm7, %ymm1
+; AVX512DQ-NEXT: vpshufb %ymm6, %ymm7, %ymm1
; AVX512DQ-NEXT: vpmovsxbd {{.*#+}} ymm2 = [0,0,0,0,1,1,0,2]
; AVX512DQ-NEXT: vpermd %ymm7, %ymm2, %ymm2
-; AVX512DQ-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512DQ-NEXT: vpandn %ymm2, %ymm3, %ymm2
+; AVX512DQ-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm2
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 | (zmm0 & zmm3)
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 | (zmm0 & mem)
; AVX512DQ-NEXT: vmovdqa64 %zmm1, (%rcx)
; AVX512DQ-NEXT: vmovdqa64 %zmm4, 128(%rcx)
; AVX512DQ-NEXT: vmovdqa64 %zmm5, 64(%rcx)
@@ -1897,50 +1897,50 @@ define void @store_i16_stride3_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm4[0,1,2,3]
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm4 = [5,5,0,6,6,0,7,7,0,8,8,0,9,9,0,10]
; AVX512DQ-FCP-NEXT: vpermd (%rdx), %zmm4, %zmm5
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (mem & (zmm5 ^ zmm3))
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm6 = [0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (zmm6 & (zmm5 ^ zmm3))
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %ymm3
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128,20,21,128,128]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rsi), %ymm8
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm9 = [10,11,0,1,128,128,12,13,2,3,128,128,14,15,4,5,128,128,16,17,28,29,128,128,18,19,18,19,128,128,20,21]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm8, %ymm8
-; AVX512DQ-FCP-NEXT: vpor %ymm3, %ymm8, %ymm3
-; AVX512DQ-FCP-NEXT: vmovdqa 48(%rdi), %xmm8
-; AVX512DQ-FCP-NEXT: vmovdqa 48(%rsi), %xmm10
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm11 = xmm10[4],xmm8[4],xmm10[5],xmm8[5],xmm10[6],xmm8[6],xmm10[7],xmm8[7]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm7, %xmm11, %xmm7
-; AVX512DQ-FCP-NEXT: vprold $16, %xmm10, %xmm10
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[1,1,2,2]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1],xmm10[2],xmm8[3,4],xmm10[5],xmm8[6,7]
-; AVX512DQ-FCP-NEXT: vinserti128 $1, %xmm7, %ymm8, %ymm7
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm8 = [128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128,20,21,128,128]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm3, %ymm3
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rsi), %ymm9
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm10 = [10,11,0,1,128,128,12,13,2,3,128,128,14,15,4,5,128,128,16,17,28,29,128,128,18,19,18,19,128,128,20,21]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm9, %ymm9
+; AVX512DQ-FCP-NEXT: vpor %ymm3, %ymm9, %ymm3
+; AVX512DQ-FCP-NEXT: vmovdqa 48(%rdi), %xmm9
+; AVX512DQ-FCP-NEXT: vmovdqa 48(%rsi), %xmm11
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm12 = xmm11[4],xmm9[4],xmm11[5],xmm9[5],xmm11[6],xmm9[6],xmm11[7],xmm9[7]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm7, %xmm12, %xmm7
+; AVX512DQ-FCP-NEXT: vprold $16, %xmm11, %xmm11
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[1,1,2,2]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm11[2],xmm9[3,4],xmm11[5],xmm9[6,7]
+; AVX512DQ-FCP-NEXT: vinserti128 $1, %xmm7, %ymm9, %ymm7
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm7[0,1,2,3]
; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %ymm7
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %ymm8
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm10 = [128,128,10,11,128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm8, %ymm11
-; AVX512DQ-FCP-NEXT: vpermd %ymm8, %ymm4, %ymm4
-; AVX512DQ-FCP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm4
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm11, %zmm4
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %ymm9
+; AVX512DQ-FCP-NEXT: vpermd %ymm9, %ymm4, %ymm4
+; AVX512DQ-FCP-NEXT: vpandn %ymm4, %ymm6, %ymm4
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,10,11,128,128,128,128,12,13,128,128,128,128,14,15,128,128,128,128,16,17,128,128,128,128,18,19,128,128,128,128]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm9, %ymm9
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm9, %zmm4
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm4 | (zmm3 & mem)
; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %ymm3
-; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %ymm6
-; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm6, %ymm6
-; AVX512DQ-FCP-NEXT: vpor %ymm3, %ymm6, %ymm3
-; AVX512DQ-FCP-NEXT: vprold $16, %xmm0, %xmm6
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm8 = xmm1[1,1,2,2]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm6 = xmm8[0,1],xmm6[2],xmm8[3,4],xmm6[5],xmm8[6,7]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm3, %ymm3
+; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %ymm8
+; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm8, %ymm8
+; AVX512DQ-FCP-NEXT: vpor %ymm3, %ymm8, %ymm3
+; AVX512DQ-FCP-NEXT: vprold $16, %xmm0, %xmm8
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm9 = xmm1[1,1,2,2]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm8 = xmm9[0,1],xmm8[2],xmm9[3,4],xmm8[5],xmm9[6,7]
; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm0, %xmm0
-; AVX512DQ-FCP-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm0
+; AVX512DQ-FCP-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm0
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm3[0,1,2,3]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm7, %ymm1
+; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm7, %ymm1
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm2 = [0,0,0,0,1,1,0,2]
; AVX512DQ-FCP-NEXT: vpermd %ymm7, %ymm2, %ymm2
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512DQ-FCP-NEXT: vpandn %ymm2, %ymm3, %ymm2
+; AVX512DQ-FCP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm2
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 | (zmm0 & zmm3)
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 | (zmm0 & mem)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm1, (%rcx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm4, 128(%rcx)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm5, 64(%rcx)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll
index 322d606538c54..418c987ab9a30 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll
@@ -3362,54 +3362,54 @@ define void @store_i16_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
; AVX512-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
; AVX512-NEXT: vmovdqa (%rdi), %ymm4
-; AVX512-NEXT: vpshufd {{.*#+}} ymm13 = ymm4[2,3,2,3,6,7,6,7]
+; AVX512-NEXT: vpshufd {{.*#+}} ymm12 = ymm4[2,3,2,3,6,7,6,7]
; AVX512-NEXT: vmovdqa (%rsi), %ymm6
; AVX512-NEXT: vpshufhw {{.*#+}} ymm14 = ymm6[0,1,2,3,7,6,5,7,8,9,10,11,15,14,13,15]
; AVX512-NEXT: vpshufd {{.*#+}} ymm14 = ymm14[2,3,2,2,6,7,6,6]
-; AVX512-NEXT: vpblendw {{.*#+}} ymm13 = ymm14[0],ymm13[1],ymm14[2],ymm13[3],ymm14[4,5],ymm13[6],ymm14[7,8],ymm13[9],ymm14[10],ymm13[11],ymm14[12,13],ymm13[14],ymm14[15]
-; AVX512-NEXT: vpermq {{.*#+}} ymm13 = ymm13[2,3,2,2]
-; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm13 = zmm13[0,1,2,3],zmm2[0,1,0,1]
+; AVX512-NEXT: vpblendw {{.*#+}} ymm12 = ymm14[0],ymm12[1],ymm14[2],ymm12[3],ymm14[4,5],ymm12[6],ymm14[7,8],ymm12[9],ymm14[10],ymm12[11],ymm14[12,13],ymm12[14],ymm14[15]
+; AVX512-NEXT: vpermq {{.*#+}} ymm12 = ymm12[2,3,2,2]
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm2[0,1,0,1]
; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm17 = [65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535]
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm13 = zmm0 ^ (zmm17 & (zmm13 ^ zmm0))
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm12 = zmm0 ^ (zmm17 & (zmm12 ^ zmm0))
; AVX512-NEXT: vpbroadcastq 24(%r8), %ymm0
; AVX512-NEXT: vpbroadcastq 32(%r8), %ymm2
; AVX512-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm19
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm19 = zmm19 ^ (mem & (zmm19 ^ zmm13))
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm19 = zmm19 ^ (mem & (zmm19 ^ zmm12))
; AVX512-NEXT: vmovdqa 32(%rsi), %ymm14
; AVX512-NEXT: vmovdqa {{.*#+}} ymm2 = [u,u,0,1,u,u,u,u,14,15,u,u,2,3,u,u,u,u,16,17,u,u,u,u,30,31,u,u,18,19,u,u]
; AVX512-NEXT: vpshufb %ymm2, %ymm14, %ymm0
; AVX512-NEXT: vmovdqa64 %ymm2, %ymm21
; AVX512-NEXT: vmovdqa64 32(%rdi), %ymm16
-; AVX512-NEXT: vpermq {{.*#+}} ymm13 = ymm16[1,1,2,2]
-; AVX512-NEXT: vpblendw {{.*#+}} ymm13 = ymm13[0],ymm0[1],ymm13[2,3],ymm0[4],ymm13[5],ymm0[6],ymm13[7,8],ymm0[9],ymm13[10,11],ymm0[12],ymm13[13],ymm0[14],ymm13[15]
+; AVX512-NEXT: vpermq {{.*#+}} ymm12 = ymm16[1,1,2,2]
+; AVX512-NEXT: vpblendw {{.*#+}} ymm12 = ymm12[0],ymm0[1],ymm12[2,3],ymm0[4],ymm12[5],ymm0[6],ymm12[7,8],ymm0[9],ymm12[10,11],ymm0[12],ymm12[13],ymm0[14],ymm12[15]
; AVX512-NEXT: vmovdqa {{.*#+}} xmm0 = [6,7,u,u,10,11,6,7,u,u,8,9,u,u,12,13]
; AVX512-NEXT: vpshufb %xmm0, %xmm1, %xmm1
; AVX512-NEXT: vpbroadcastq 40(%rdi), %xmm15
; AVX512-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm15[1],xmm1[2,3],xmm15[4],xmm1[5],xmm15[6],xmm1[7]
; AVX512-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
-; AVX512-NEXT: vinserti64x4 $1, %ymm13, %zmm1, %zmm13
+; AVX512-NEXT: vinserti64x4 $1, %ymm12, %zmm1, %zmm12
; AVX512-NEXT: vmovdqa 32(%rcx), %ymm15
; AVX512-NEXT: vmovdqa {{.*#+}} ymm2 = [u,u,12,13,u,u,0,1,u,u,u,u,14,15,u,u,u,u,28,29,u,u,16,17,u,u,u,u,30,31,u,u]
; AVX512-NEXT: vpshufb %ymm2, %ymm15, %ymm1
; AVX512-NEXT: vmovdqa64 %ymm2, %ymm22
-; AVX512-NEXT: vpshufd {{.*#+}} ymm12 = ymm5[3,0,3,0,7,4,7,4]
-; AVX512-NEXT: vpblendw {{.*#+}} ymm12 = ymm12[0],ymm1[1],ymm12[2],ymm1[3],ymm12[4,5],ymm1[6],ymm12[7,8],ymm1[9],ymm12[10],ymm1[11],ymm12[12,13],ymm1[14],ymm12[15]
+; AVX512-NEXT: vpshufd {{.*#+}} ymm13 = ymm5[3,0,3,0,7,4,7,4]
+; AVX512-NEXT: vpblendw {{.*#+}} ymm13 = ymm13[0],ymm1[1],ymm13[2],ymm1[3],ymm13[4,5],ymm1[6],ymm13[7,8],ymm1[9],ymm13[10],ymm1[11],ymm13[12,13],ymm1[14],ymm13[15]
; AVX512-NEXT: vmovdqa {{.*#+}} xmm1 = [10,11,u,u,6,7,u,u,8,9,8,9,u,u,8,9]
; AVX512-NEXT: vpshufb %xmm1, %xmm11, %xmm11
; AVX512-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[1,2,2,2]
; AVX512-NEXT: vpblendw {{.*#+}} xmm10 = xmm11[0],xmm10[1],xmm11[2],xmm10[3],xmm11[4,5],xmm10[6],xmm11[7]
; AVX512-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,1,0,0]
-; AVX512-NEXT: vinserti64x4 $1, %ymm12, %zmm10, %zmm12
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm12 = zmm12 ^ (mem & (zmm12 ^ zmm13))
-; AVX512-NEXT: vmovdqa (%r8), %ymm11
-; AVX512-NEXT: vmovdqa 32(%r8), %ymm10
-; AVX512-NEXT: vmovdqa {{.*#+}} ymm13 = [128,128,128,128,12,13,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128]
-; AVX512-NEXT: vpshufb %ymm13, %ymm10, %ymm2
+; AVX512-NEXT: vinserti64x4 $1, %ymm13, %zmm10, %zmm13
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm13 = zmm13 ^ (mem & (zmm13 ^ zmm12))
+; AVX512-NEXT: vmovdqa (%r8), %ymm10
+; AVX512-NEXT: vmovdqa 32(%r8), %ymm11
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm12 = [128,128,128,128,12,13,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128]
+; AVX512-NEXT: vpshufb %ymm12, %ymm11, %ymm2
; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm20 = [65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535]
-; AVX512-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,1,1,1]
-; AVX512-NEXT: vpandnq %ymm10, %ymm20, %ymm10
-; AVX512-NEXT: vinserti64x4 $1, %ymm2, %zmm10, %zmm10
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm10 = zmm10 | (zmm12 & zmm20)
+; AVX512-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,1,1,1]
+; AVX512-NEXT: vpandnq %ymm11, %ymm20, %ymm11
+; AVX512-NEXT: vinserti64x4 $1, %ymm2, %zmm11, %zmm11
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm11 = zmm11 | (zmm13 & zmm20)
; AVX512-NEXT: vpshufb %xmm0, %xmm3, %xmm0
; AVX512-NEXT: vpbroadcastq 8(%rdi), %xmm2
; AVX512-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3],xmm2[4],xmm0[5],xmm2[6],xmm0[7]
@@ -3430,7 +3430,7 @@ define void @store_i16_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535]
; AVX512-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (zmm2 & (zmm1 ^ zmm0))
; AVX512-NEXT: vpbroadcastq (%r8), %ymm0
-; AVX512-NEXT: vpermq {{.*#+}} ymm3 = ymm11[0,1,1,1]
+; AVX512-NEXT: vpermq {{.*#+}} ymm3 = ymm10[0,1,1,1]
; AVX512-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
; AVX512-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm1))
; AVX512-NEXT: vpshufd {{.*#+}} ymm1 = ymm16[0,1,2,1,4,5,6,5]
@@ -3459,33 +3459,33 @@ define void @store_i16_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-NEXT: vpbroadcastq 48(%r8), %ymm1
; AVX512-NEXT: vpbroadcastq 56(%r8), %ymm2
; AVX512-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (mem & (zmm1 ^ zmm5))
-; AVX512-NEXT: vpshufd {{.*#+}} ymm2 = ymm4[0,1,2,1,4,5,6,5]
-; AVX512-NEXT: vprolq $16, %ymm6, %ymm5
-; AVX512-NEXT: vpblendw {{.*#+}} ymm2 = ymm5[0,1],ymm2[2],ymm5[3],ymm2[4],ymm5[5,6],ymm2[7],ymm5[8,9],ymm2[10],ymm5[11],ymm2[12],ymm5[13,14],ymm2[15]
-; AVX512-NEXT: vmovdqa64 %ymm21, %ymm5
-; AVX512-NEXT: vpshufb %ymm5, %ymm6, %ymm5
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0]
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (zmm2 & (zmm1 ^ zmm5))
+; AVX512-NEXT: vpshufd {{.*#+}} ymm5 = ymm4[0,1,2,1,4,5,6,5]
+; AVX512-NEXT: vprolq $16, %ymm6, %ymm8
+; AVX512-NEXT: vpblendw {{.*#+}} ymm5 = ymm8[0,1],ymm5[2],ymm8[3],ymm5[4],ymm8[5,6],ymm5[7],ymm8[8,9],ymm5[10],ymm8[11],ymm5[12],ymm8[13,14],ymm5[15]
+; AVX512-NEXT: vmovdqa64 %ymm21, %ymm8
+; AVX512-NEXT: vpshufb %ymm8, %ymm6, %ymm6
; AVX512-NEXT: vpermq {{.*#+}} ymm4 = ymm4[1,1,2,2]
-; AVX512-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3],ymm5[4],ymm4[5],ymm5[6],ymm4[7,8],ymm5[9],ymm4[10,11],ymm5[12],ymm4[13],ymm5[14],ymm4[15]
-; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm4[0,1,2,3],zmm2[2,3,2,3]
+; AVX512-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm6[1],ymm4[2,3],ymm6[4],ymm4[5],ymm6[6],ymm4[7,8],ymm6[9],ymm4[10,11],ymm6[12],ymm4[13],ymm6[14],ymm4[15]
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[2,3,2,3]
; AVX512-NEXT: vpshufb %ymm3, %ymm7, %ymm3
-; AVX512-NEXT: vpshufd {{.*#+}} ymm4 = ymm18[1,1,1,2,5,5,5,6]
-; AVX512-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5],ymm4[6],ymm3[7,8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13],ymm4[14],ymm3[15]
-; AVX512-NEXT: vmovdqa64 %ymm22, %ymm4
-; AVX512-NEXT: vpshufb %ymm4, %ymm7, %ymm4
-; AVX512-NEXT: vpshufd {{.*#+}} ymm5 = ymm18[3,0,3,0,7,4,7,4]
-; AVX512-NEXT: vpblendw {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2],ymm4[3],ymm5[4,5],ymm4[6],ymm5[7,8],ymm4[9],ymm5[10],ymm4[11],ymm5[12,13],ymm4[14],ymm5[15]
-; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm4[0,1,2,3],zmm3[2,3,2,3]
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm3 = zmm2 ^ (zmm17 & (zmm3 ^ zmm2))
-; AVX512-NEXT: vpshufb %ymm13, %ymm11, %ymm2
-; AVX512-NEXT: vpbroadcastq 16(%r8), %ymm4
-; AVX512-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm4
-; AVX512-NEXT: vinserti64x4 $1, %ymm4, %zmm2, %zmm2
+; AVX512-NEXT: vpshufd {{.*#+}} ymm5 = ymm18[1,1,1,2,5,5,5,6]
+; AVX512-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2,3],ymm5[4],ymm3[5],ymm5[6],ymm3[7,8],ymm5[9],ymm3[10,11],ymm5[12],ymm3[13],ymm5[14],ymm3[15]
+; AVX512-NEXT: vmovdqa64 %ymm22, %ymm5
+; AVX512-NEXT: vpshufb %ymm5, %ymm7, %ymm5
+; AVX512-NEXT: vpshufd {{.*#+}} ymm6 = ymm18[3,0,3,0,7,4,7,4]
+; AVX512-NEXT: vpblendw {{.*#+}} ymm5 = ymm6[0],ymm5[1],ymm6[2],ymm5[3],ymm6[4,5],ymm5[6],ymm6[7,8],ymm5[9],ymm6[10],ymm5[11],ymm6[12,13],ymm5[14],ymm6[15]
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm5[0,1,2,3],zmm3[2,3,2,3]
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm3 = zmm4 ^ (zmm17 & (zmm3 ^ zmm4))
+; AVX512-NEXT: vpandnq 16(%r8){1to4}, %ymm2, %ymm2
+; AVX512-NEXT: vpshufb %ymm12, %ymm10, %ymm4
+; AVX512-NEXT: vinserti64x4 $1, %ymm2, %zmm4, %zmm2
; AVX512-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 | (zmm3 & mem)
; AVX512-NEXT: vmovdqa64 %zmm2, 64(%r9)
; AVX512-NEXT: vmovdqa64 %zmm1, 256(%r9)
; AVX512-NEXT: vmovdqa64 %zmm0, (%r9)
-; AVX512-NEXT: vmovdqa64 %zmm10, 192(%r9)
+; AVX512-NEXT: vmovdqa64 %zmm11, 192(%r9)
; AVX512-NEXT: vmovdqa64 %zmm19, 128(%r9)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
@@ -3556,81 +3556,81 @@ define void @store_i16_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm4[2,3,2,3]
; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm17 = [65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535]
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm2 ^ (zmm17 & (zmm1 ^ zmm2))
-; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm6, %ymm2
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0]
+; AVX512-FCP-NEXT: vpandnq 16(%r8){1to4}, %ymm2, %ymm4
+; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm6, %ymm5
; AVX512-FCP-NEXT: vmovdqa64 %ymm6, %ymm20
-; AVX512-FCP-NEXT: vpbroadcastq 16(%r8), %ymm4
-; AVX512-FCP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm4
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm2, %zmm7
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm5, %zmm7
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 | (zmm1 & mem)
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [30,31,28,29,26,27,30,31,30,31,28,29,30,31,28,29,30,31,28,29,26,27,30,31,30,31,28,29,30,31,28,29]
; AVX512-FCP-NEXT: # ymm1 = mem[0,1,0,1]
-; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm10, %ymm2
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm13[2,3,2,3,6,7,6,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[3],ymm2[4,5],ymm4[6],ymm2[7,8],ymm4[9],ymm2[10],ymm4[11],ymm2[12,13],ymm4[14],ymm2[15]
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm13[0,1,2,1,4,5,6,5]
-; AVX512-FCP-NEXT: vprolq $16, %ymm10, %ymm5
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3],ymm4[4],ymm5[5,6],ymm4[7],ymm5[8,9],ymm4[10],ymm5[11],ymm4[12],ymm5[13,14],ymm4[15]
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm5 = [2,3,2,3,10,11,10,10]
-; AVX512-FCP-NEXT: vpermi2q %zmm2, %zmm4, %zmm5
-; AVX512-FCP-NEXT: vpbroadcastq {{.*#+}} ymm2 = [26,27,28,29,30,31,30,31,26,27,28,29,30,31,30,31,26,27,28,29,30,31,30,31,26,27,28,29,30,31,30,31]
-; AVX512-FCP-NEXT: vpshufb %ymm2, %ymm14, %ymm4
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm10 = ymm16[3,2,3,3,7,6,7,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm10[1],ymm4[2],ymm10[3,4],ymm4[5,6,7,8],ymm10[9],ymm4[10],ymm10[11,12],ymm4[13,14,15]
+; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm10, %ymm4
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm13[2,3,2,3,6,7,6,7]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2],ymm5[3],ymm4[4,5],ymm5[6],ymm4[7,8],ymm5[9],ymm4[10],ymm5[11],ymm4[12,13],ymm5[14],ymm4[15]
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm13[0,1,2,1,4,5,6,5]
+; AVX512-FCP-NEXT: vprolq $16, %ymm10, %ymm10
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm5 = ymm10[0,1],ymm5[2],ymm10[3],ymm5[4],ymm10[5,6],ymm5[7],ymm10[8,9],ymm5[10],ymm10[11],ymm5[12],ymm10[13,14],ymm5[15]
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm10 = [2,3,2,3,10,11,10,10]
+; AVX512-FCP-NEXT: vpermi2q %zmm4, %zmm5, %zmm10
+; AVX512-FCP-NEXT: vpbroadcastq {{.*#+}} ymm4 = [26,27,28,29,30,31,30,31,26,27,28,29,30,31,30,31,26,27,28,29,30,31,30,31,26,27,28,29,30,31,30,31]
+; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm14, %ymm5
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm13 = ymm16[3,2,3,3,7,6,7,7]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0],ymm13[1],ymm5[2],ymm13[3,4],ymm5[5,6,7,8],ymm13[9],ymm5[10],ymm13[11,12],ymm5[13,14,15]
; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm14, %ymm0
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm10 = ymm16[1,1,1,2,5,5,5,6]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm10[1],ymm0[2,3],ymm10[4],ymm0[5],ymm10[6],ymm0[7,8],ymm10[9],ymm0[10,11],ymm10[12],ymm0[13],ymm10[14],ymm0[15]
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm13 = ymm16[1,1,1,2,5,5,5,6]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm13[1],ymm0[2,3],ymm13[4],ymm0[5],ymm13[6],ymm0[7,8],ymm13[9],ymm0[10,11],ymm13[12],ymm0[13],ymm13[14],ymm0[15]
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm13 = [2,3,2,3,10,10,11,10]
-; AVX512-FCP-NEXT: vpermi2q %zmm4, %zmm0, %zmm13
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm4 = [65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm13 = zmm5 ^ (zmm4 & (zmm13 ^ zmm5))
+; AVX512-FCP-NEXT: vpermi2q %zmm5, %zmm0, %zmm13
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm14 = [65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm13 = zmm10 ^ (zmm14 & (zmm13 ^ zmm10))
; AVX512-FCP-NEXT: vpbroadcastq 48(%r8), %ymm0
; AVX512-FCP-NEXT: vpbroadcastq 56(%r8), %ymm5
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm5, %zmm0, %zmm10
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm10 = zmm10 ^ (mem & (zmm10 ^ zmm13))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm10 = zmm10 ^ (zmm2 & (zmm10 ^ zmm13))
; AVX512-FCP-NEXT: vmovdqa64 %xmm23, %xmm0
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm13 = [4,5,2,3,2,3,0,1,10,11,8,9,4,5,6,7]
-; AVX512-FCP-NEXT: vpshufb %xmm13, %xmm0, %xmm0
-; AVX512-FCP-NEXT: vpshufb %ymm2, %ymm9, %ymm2
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm0 = [4,5,2,3,2,3,0,1,10,11,8,9,4,5,6,7]
+; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm2, %xmm2
+; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm9, %ymm4
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm8[3,2,3,3,7,6,7,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2],ymm5[3,4],ymm2[5,6,7,8],ymm5[9],ymm2[10],ymm5[11,12],ymm2[13,14,15]
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm6 = [2,2,3,2,8,9,8,9]
-; AVX512-FCP-NEXT: vpermi2q %zmm0, %zmm2, %zmm6
-; AVX512-FCP-NEXT: vmovdqa (%rdi), %xmm0
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2],ymm5[3,4],ymm4[5,6,7,8],ymm5[9],ymm4[10],ymm5[11,12],ymm4[13,14,15]
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm8 = [2,2,3,2,8,9,8,9]
+; AVX512-FCP-NEXT: vpermi2q %zmm2, %zmm4, %zmm8
+; AVX512-FCP-NEXT: vmovdqa (%rdi), %xmm6
; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %xmm2
-; AVX512-FCP-NEXT: vmovdqa64 %xmm22, %xmm5
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
+; AVX512-FCP-NEXT: vmovdqa64 %xmm22, %xmm4
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm5 = [0,1,2,3,8,9,10,11,4,5,4,5,6,7,12,13]
; AVX512-FCP-NEXT: vpshufb %xmm5, %xmm2, %xmm2
; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm15, %ymm1
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm8 = ymm12[2,3,2,3,6,7,6,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7,8],ymm8[9],ymm1[10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm8 = [2,3,2,2,8,9,8,9]
-; AVX512-FCP-NEXT: vpermi2q %zmm2, %zmm1, %zmm8
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm6 ^ (zmm17 & (zmm8 ^ zmm6))
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm12[2,3,2,3,6,7,6,7]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2],ymm4[3],ymm1[4,5],ymm4[6],ymm1[7,8],ymm4[9],ymm1[10],ymm4[11],ymm1[12,13],ymm4[14],ymm1[15]
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm4 = [2,3,2,2,8,9,8,9]
+; AVX512-FCP-NEXT: vpermi2q %zmm2, %zmm1, %zmm4
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm8 ^ (zmm17 & (zmm4 ^ zmm8))
; AVX512-FCP-NEXT: vpbroadcastq 24(%r8), %ymm1
; AVX512-FCP-NEXT: vpbroadcastq 32(%r8), %ymm2
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (mem & (zmm1 ^ zmm8))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (mem & (zmm1 ^ zmm4))
; AVX512-FCP-NEXT: vmovdqa (%rcx), %xmm2
-; AVX512-FCP-NEXT: vmovdqa64 %xmm21, %xmm6
-; AVX512-FCP-NEXT: vpshufb %xmm6, %xmm2, %xmm6
+; AVX512-FCP-NEXT: vmovdqa64 %xmm21, %xmm4
+; AVX512-FCP-NEXT: vpshufb %xmm4, %xmm2, %xmm4
; AVX512-FCP-NEXT: vmovdqa (%rdx), %xmm8
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm11 = xmm8[1,2,2,2]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0],xmm11[1],xmm6[2],xmm11[3],xmm6[4,5],xmm11[6],xmm6[7]
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm9 = xmm8[1,2,2,2]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0],xmm9[1],xmm4[2],xmm9[3],xmm4[4,5],xmm9[6],xmm4[7]
; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
-; AVX512-FCP-NEXT: vpshufb %xmm13, %xmm2, %xmm2
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm8 = [0,1,0,1,8,9,8,8]
-; AVX512-FCP-NEXT: vpermi2q %zmm6, %zmm2, %zmm8
-; AVX512-FCP-NEXT: vmovdqa64 %xmm18, %xmm9
-; AVX512-FCP-NEXT: vmovdqa64 %xmm19, %xmm2
-; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm9, %xmm2
-; AVX512-FCP-NEXT: vpbroadcastq 8(%rdi), %xmm6
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm6[1],xmm2[2,3],xmm6[4],xmm2[5],xmm6[6],xmm2[7]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3]
-; AVX512-FCP-NEXT: vpshufb %xmm5, %xmm0, %xmm0
-; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,0,1],zmm2[0,1,0,1]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm8 ^ (zmm4 & (zmm0 ^ zmm8))
+; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm2, %xmm0
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm2 = [0,1,0,1,8,9,8,8]
+; AVX512-FCP-NEXT: vpermi2q %zmm4, %zmm0, %zmm2
+; AVX512-FCP-NEXT: vmovdqa64 %xmm18, %xmm8
+; AVX512-FCP-NEXT: vmovdqa64 %xmm19, %xmm0
+; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm8, %xmm0
+; AVX512-FCP-NEXT: vpbroadcastq 8(%rdi), %xmm4
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm4[1],xmm0[2,3],xmm4[4],xmm0[5],xmm4[6],xmm0[7]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm8[0],xmm6[1],xmm8[1],xmm6[2],xmm8[2],xmm6[3],xmm8[3]
+; AVX512-FCP-NEXT: vpshufb %xmm5, %xmm4, %xmm4
+; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm4[0,1,0,1],zmm0[0,1,0,1]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm2 ^ (zmm14 & (zmm0 ^ zmm2))
; AVX512-FCP-NEXT: vpbroadcastq (%r8), %ymm2
; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm4 = ymm20[0,1,1,1]
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm2, %zmm2
@@ -3669,54 +3669,54 @@ define void @store_i16_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
; AVX512DQ-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
; AVX512DQ-NEXT: vmovdqa (%rdi), %ymm4
-; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm13 = ymm4[2,3,2,3,6,7,6,7]
+; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm12 = ymm4[2,3,2,3,6,7,6,7]
; AVX512DQ-NEXT: vmovdqa (%rsi), %ymm6
; AVX512DQ-NEXT: vpshufhw {{.*#+}} ymm14 = ymm6[0,1,2,3,7,6,5,7,8,9,10,11,15,14,13,15]
; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm14 = ymm14[2,3,2,2,6,7,6,6]
-; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm13 = ymm14[0],ymm13[1],ymm14[2],ymm13[3],ymm14[4,5],ymm13[6],ymm14[7,8],ymm13[9],ymm14[10],ymm13[11],ymm14[12,13],ymm13[14],ymm14[15]
-; AVX512DQ-NEXT: vpermq {{.*#+}} ymm13 = ymm13[2,3,2,2]
-; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm13 = zmm13[0,1,2,3],zmm2[0,1,0,1]
+; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm12 = ymm14[0],ymm12[1],ymm14[2],ymm12[3],ymm14[4,5],ymm12[6],ymm14[7,8],ymm12[9],ymm14[10],ymm12[11],ymm14[12,13],ymm12[14],ymm14[15]
+; AVX512DQ-NEXT: vpermq {{.*#+}} ymm12 = ymm12[2,3,2,2]
+; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm2[0,1,0,1]
; AVX512DQ-NEXT: vmovdqa64 {{.*#+}} zmm17 = [65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535]
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm13 = zmm0 ^ (zmm17 & (zmm13 ^ zmm0))
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm12 = zmm0 ^ (zmm17 & (zmm12 ^ zmm0))
; AVX512DQ-NEXT: vpbroadcastq 24(%r8), %ymm0
; AVX512DQ-NEXT: vpbroadcastq 32(%r8), %ymm2
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm19
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm19 = zmm19 ^ (mem & (zmm19 ^ zmm13))
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm19 = zmm19 ^ (mem & (zmm19 ^ zmm12))
; AVX512DQ-NEXT: vmovdqa 32(%rsi), %ymm14
; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm2 = [u,u,0,1,u,u,u,u,14,15,u,u,2,3,u,u,u,u,16,17,u,u,u,u,30,31,u,u,18,19,u,u]
; AVX512DQ-NEXT: vpshufb %ymm2, %ymm14, %ymm0
; AVX512DQ-NEXT: vmovdqa64 %ymm2, %ymm21
; AVX512DQ-NEXT: vmovdqa64 32(%rdi), %ymm16
-; AVX512DQ-NEXT: vpermq {{.*#+}} ymm13 = ymm16[1,1,2,2]
-; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm13 = ymm13[0],ymm0[1],ymm13[2,3],ymm0[4],ymm13[5],ymm0[6],ymm13[7,8],ymm0[9],ymm13[10,11],ymm0[12],ymm13[13],ymm0[14],ymm13[15]
+; AVX512DQ-NEXT: vpermq {{.*#+}} ymm12 = ymm16[1,1,2,2]
+; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm12 = ymm12[0],ymm0[1],ymm12[2,3],ymm0[4],ymm12[5],ymm0[6],ymm12[7,8],ymm0[9],ymm12[10,11],ymm0[12],ymm12[13],ymm0[14],ymm12[15]
; AVX512DQ-NEXT: vmovdqa {{.*#+}} xmm0 = [6,7,u,u,10,11,6,7,u,u,8,9,u,u,12,13]
; AVX512DQ-NEXT: vpshufb %xmm0, %xmm1, %xmm1
; AVX512DQ-NEXT: vpbroadcastq 40(%rdi), %xmm15
; AVX512DQ-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm15[1],xmm1[2,3],xmm15[4],xmm1[5],xmm15[6],xmm1[7]
; AVX512DQ-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm13, %zmm1, %zmm13
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm12, %zmm1, %zmm12
; AVX512DQ-NEXT: vmovdqa 32(%rcx), %ymm15
; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm2 = [u,u,12,13,u,u,0,1,u,u,u,u,14,15,u,u,u,u,28,29,u,u,16,17,u,u,u,u,30,31,u,u]
; AVX512DQ-NEXT: vpshufb %ymm2, %ymm15, %ymm1
; AVX512DQ-NEXT: vmovdqa64 %ymm2, %ymm22
-; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm12 = ymm5[3,0,3,0,7,4,7,4]
-; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm12 = ymm12[0],ymm1[1],ymm12[2],ymm1[3],ymm12[4,5],ymm1[6],ymm12[7,8],ymm1[9],ymm12[10],ymm1[11],ymm12[12,13],ymm1[14],ymm12[15]
+; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm13 = ymm5[3,0,3,0,7,4,7,4]
+; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm13 = ymm13[0],ymm1[1],ymm13[2],ymm1[3],ymm13[4,5],ymm1[6],ymm13[7,8],ymm1[9],ymm13[10],ymm1[11],ymm13[12,13],ymm1[14],ymm13[15]
; AVX512DQ-NEXT: vmovdqa {{.*#+}} xmm1 = [10,11,u,u,6,7,u,u,8,9,8,9,u,u,8,9]
; AVX512DQ-NEXT: vpshufb %xmm1, %xmm11, %xmm11
; AVX512DQ-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[1,2,2,2]
; AVX512DQ-NEXT: vpblendw {{.*#+}} xmm10 = xmm11[0],xmm10[1],xmm11[2],xmm10[3],xmm11[4,5],xmm10[6],xmm11[7]
; AVX512DQ-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,1,0,0]
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm12, %zmm10, %zmm12
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm12 = zmm12 ^ (mem & (zmm12 ^ zmm13))
-; AVX512DQ-NEXT: vmovdqa (%r8), %ymm11
-; AVX512DQ-NEXT: vmovdqa 32(%r8), %ymm10
-; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm13 = [128,128,128,128,12,13,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128]
-; AVX512DQ-NEXT: vpshufb %ymm13, %ymm10, %ymm2
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm13, %zmm10, %zmm13
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm13 = zmm13 ^ (mem & (zmm13 ^ zmm12))
+; AVX512DQ-NEXT: vmovdqa (%r8), %ymm10
+; AVX512DQ-NEXT: vmovdqa 32(%r8), %ymm11
+; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm12 = [128,128,128,128,12,13,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128]
+; AVX512DQ-NEXT: vpshufb %ymm12, %ymm11, %ymm2
; AVX512DQ-NEXT: vmovdqa64 {{.*#+}} zmm20 = [65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535]
-; AVX512DQ-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,1,1,1]
-; AVX512DQ-NEXT: vpandnq %ymm10, %ymm20, %ymm10
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm2, %zmm10, %zmm10
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm10 = zmm10 | (zmm12 & zmm20)
+; AVX512DQ-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,1,1,1]
+; AVX512DQ-NEXT: vpandnq %ymm11, %ymm20, %ymm11
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm2, %zmm11, %zmm11
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm11 = zmm11 | (zmm13 & zmm20)
; AVX512DQ-NEXT: vpshufb %xmm0, %xmm3, %xmm0
; AVX512DQ-NEXT: vpbroadcastq 8(%rdi), %xmm2
; AVX512DQ-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3],xmm2[4],xmm0[5],xmm2[6],xmm0[7]
@@ -3737,7 +3737,7 @@ define void @store_i16_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535]
; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (zmm2 & (zmm1 ^ zmm0))
; AVX512DQ-NEXT: vpbroadcastq (%r8), %ymm0
-; AVX512DQ-NEXT: vpermq {{.*#+}} ymm3 = ymm11[0,1,1,1]
+; AVX512DQ-NEXT: vpermq {{.*#+}} ymm3 = ymm10[0,1,1,1]
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm1))
; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm1 = ymm16[0,1,2,1,4,5,6,5]
@@ -3766,33 +3766,33 @@ define void @store_i16_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-NEXT: vpbroadcastq 48(%r8), %ymm1
; AVX512DQ-NEXT: vpbroadcastq 56(%r8), %ymm2
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (mem & (zmm1 ^ zmm5))
-; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm2 = ymm4[0,1,2,1,4,5,6,5]
-; AVX512DQ-NEXT: vprolq $16, %ymm6, %ymm5
-; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm2 = ymm5[0,1],ymm2[2],ymm5[3],ymm2[4],ymm5[5,6],ymm2[7],ymm5[8,9],ymm2[10],ymm5[11],ymm2[12],ymm5[13,14],ymm2[15]
-; AVX512DQ-NEXT: vmovdqa64 %ymm21, %ymm5
-; AVX512DQ-NEXT: vpshufb %ymm5, %ymm6, %ymm5
+; AVX512DQ-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0]
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (zmm2 & (zmm1 ^ zmm5))
+; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm5 = ymm4[0,1,2,1,4,5,6,5]
+; AVX512DQ-NEXT: vprolq $16, %ymm6, %ymm8
+; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm5 = ymm8[0,1],ymm5[2],ymm8[3],ymm5[4],ymm8[5,6],ymm5[7],ymm8[8,9],ymm5[10],ymm8[11],ymm5[12],ymm8[13,14],ymm5[15]
+; AVX512DQ-NEXT: vmovdqa64 %ymm21, %ymm8
+; AVX512DQ-NEXT: vpshufb %ymm8, %ymm6, %ymm6
; AVX512DQ-NEXT: vpermq {{.*#+}} ymm4 = ymm4[1,1,2,2]
-; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3],ymm5[4],ymm4[5],ymm5[6],ymm4[7,8],ymm5[9],ymm4[10,11],ymm5[12],ymm4[13],ymm5[14],ymm4[15]
-; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm4[0,1,2,3],zmm2[2,3,2,3]
+; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm6[1],ymm4[2,3],ymm6[4],ymm4[5],ymm6[6],ymm4[7,8],ymm6[9],ymm4[10,11],ymm6[12],ymm4[13],ymm6[14],ymm4[15]
+; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[2,3,2,3]
; AVX512DQ-NEXT: vpshufb %ymm3, %ymm7, %ymm3
-; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm4 = ymm18[1,1,1,2,5,5,5,6]
-; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5],ymm4[6],ymm3[7,8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13],ymm4[14],ymm3[15]
-; AVX512DQ-NEXT: vmovdqa64 %ymm22, %ymm4
-; AVX512DQ-NEXT: vpshufb %ymm4, %ymm7, %ymm4
-; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm5 = ymm18[3,0,3,0,7,4,7,4]
-; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2],ymm4[3],ymm5[4,5],ymm4[6],ymm5[7,8],ymm4[9],ymm5[10],ymm4[11],ymm5[12,13],ymm4[14],ymm5[15]
-; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm4[0,1,2,3],zmm3[2,3,2,3]
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm3 = zmm2 ^ (zmm17 & (zmm3 ^ zmm2))
-; AVX512DQ-NEXT: vpshufb %ymm13, %ymm11, %ymm2
-; AVX512DQ-NEXT: vpbroadcastq 16(%r8), %ymm4
-; AVX512DQ-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm4
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm4, %zmm2, %zmm2
+; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm5 = ymm18[1,1,1,2,5,5,5,6]
+; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2,3],ymm5[4],ymm3[5],ymm5[6],ymm3[7,8],ymm5[9],ymm3[10,11],ymm5[12],ymm3[13],ymm5[14],ymm3[15]
+; AVX512DQ-NEXT: vmovdqa64 %ymm22, %ymm5
+; AVX512DQ-NEXT: vpshufb %ymm5, %ymm7, %ymm5
+; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm6 = ymm18[3,0,3,0,7,4,7,4]
+; AVX512DQ-NEXT: vpblendw {{.*#+}} ymm5 = ymm6[0],ymm5[1],ymm6[2],ymm5[3],ymm6[4,5],ymm5[6],ymm6[7,8],ymm5[9],ymm6[10],ymm5[11],ymm6[12,13],ymm5[14],ymm6[15]
+; AVX512DQ-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm5[0,1,2,3],zmm3[2,3,2,3]
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm3 = zmm4 ^ (zmm17 & (zmm3 ^ zmm4))
+; AVX512DQ-NEXT: vpandnq 16(%r8){1to4}, %ymm2, %ymm2
+; AVX512DQ-NEXT: vpshufb %ymm12, %ymm10, %ymm4
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm2, %zmm4, %zmm2
; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 | (zmm3 & mem)
; AVX512DQ-NEXT: vmovdqa64 %zmm2, 64(%r9)
; AVX512DQ-NEXT: vmovdqa64 %zmm1, 256(%r9)
; AVX512DQ-NEXT: vmovdqa64 %zmm0, (%r9)
-; AVX512DQ-NEXT: vmovdqa64 %zmm10, 192(%r9)
+; AVX512DQ-NEXT: vmovdqa64 %zmm11, 192(%r9)
; AVX512DQ-NEXT: vmovdqa64 %zmm19, 128(%r9)
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
@@ -3863,81 +3863,81 @@ define void @store_i16_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm4[2,3,2,3]
; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm17 = [65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535]
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm2 ^ (zmm17 & (zmm1 ^ zmm2))
-; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm6, %ymm2
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0,65535,65535,65535,65535,0]
+; AVX512DQ-FCP-NEXT: vpandnq 16(%r8){1to4}, %ymm2, %ymm4
+; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm6, %ymm5
; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm6, %ymm20
-; AVX512DQ-FCP-NEXT: vpbroadcastq 16(%r8), %ymm4
-; AVX512DQ-FCP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm4
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm2, %zmm7
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm5, %zmm7
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 | (zmm1 & mem)
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [30,31,28,29,26,27,30,31,30,31,28,29,30,31,28,29,30,31,28,29,26,27,30,31,30,31,28,29,30,31,28,29]
; AVX512DQ-FCP-NEXT: # ymm1 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm10, %ymm2
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm13[2,3,2,3,6,7,6,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[3],ymm2[4,5],ymm4[6],ymm2[7,8],ymm4[9],ymm2[10],ymm4[11],ymm2[12,13],ymm4[14],ymm2[15]
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm13[0,1,2,1,4,5,6,5]
-; AVX512DQ-FCP-NEXT: vprolq $16, %ymm10, %ymm5
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3],ymm4[4],ymm5[5,6],ymm4[7],ymm5[8,9],ymm4[10],ymm5[11],ymm4[12],ymm5[13,14],ymm4[15]
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm5 = [2,3,2,3,10,11,10,10]
-; AVX512DQ-FCP-NEXT: vpermi2q %zmm2, %zmm4, %zmm5
-; AVX512DQ-FCP-NEXT: vpbroadcastq {{.*#+}} ymm2 = [26,27,28,29,30,31,30,31,26,27,28,29,30,31,30,31,26,27,28,29,30,31,30,31,26,27,28,29,30,31,30,31]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm2, %ymm14, %ymm4
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm10 = ymm16[3,2,3,3,7,6,7,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm10[1],ymm4[2],ymm10[3,4],ymm4[5,6,7,8],ymm10[9],ymm4[10],ymm10[11,12],ymm4[13,14,15]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm10, %ymm4
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm13[2,3,2,3,6,7,6,7]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2],ymm5[3],ymm4[4,5],ymm5[6],ymm4[7,8],ymm5[9],ymm4[10],ymm5[11],ymm4[12,13],ymm5[14],ymm4[15]
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm13[0,1,2,1,4,5,6,5]
+; AVX512DQ-FCP-NEXT: vprolq $16, %ymm10, %ymm10
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm5 = ymm10[0,1],ymm5[2],ymm10[3],ymm5[4],ymm10[5,6],ymm5[7],ymm10[8,9],ymm5[10],ymm10[11],ymm5[12],ymm10[13,14],ymm5[15]
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm10 = [2,3,2,3,10,11,10,10]
+; AVX512DQ-FCP-NEXT: vpermi2q %zmm4, %zmm5, %zmm10
+; AVX512DQ-FCP-NEXT: vpbroadcastq {{.*#+}} ymm4 = [26,27,28,29,30,31,30,31,26,27,28,29,30,31,30,31,26,27,28,29,30,31,30,31,26,27,28,29,30,31,30,31]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm14, %ymm5
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm13 = ymm16[3,2,3,3,7,6,7,7]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0],ymm13[1],ymm5[2],ymm13[3,4],ymm5[5,6,7,8],ymm13[9],ymm5[10],ymm13[11,12],ymm5[13,14,15]
; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm14, %ymm0
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm10 = ymm16[1,1,1,2,5,5,5,6]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm10[1],ymm0[2,3],ymm10[4],ymm0[5],ymm10[6],ymm0[7,8],ymm10[9],ymm0[10,11],ymm10[12],ymm0[13],ymm10[14],ymm0[15]
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm13 = ymm16[1,1,1,2,5,5,5,6]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm13[1],ymm0[2,3],ymm13[4],ymm0[5],ymm13[6],ymm0[7,8],ymm13[9],ymm0[10,11],ymm13[12],ymm0[13],ymm13[14],ymm0[15]
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm13 = [2,3,2,3,10,10,11,10]
-; AVX512DQ-FCP-NEXT: vpermi2q %zmm4, %zmm0, %zmm13
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm4 = [65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm13 = zmm5 ^ (zmm4 & (zmm13 ^ zmm5))
+; AVX512DQ-FCP-NEXT: vpermi2q %zmm5, %zmm0, %zmm13
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm14 = [65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm13 = zmm10 ^ (zmm14 & (zmm13 ^ zmm10))
; AVX512DQ-FCP-NEXT: vpbroadcastq 48(%r8), %ymm0
; AVX512DQ-FCP-NEXT: vpbroadcastq 56(%r8), %ymm5
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm5, %zmm0, %zmm10
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm10 = zmm10 ^ (mem & (zmm10 ^ zmm13))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm10 = zmm10 ^ (zmm2 & (zmm10 ^ zmm13))
; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm23, %xmm0
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm13 = [4,5,2,3,2,3,0,1,10,11,8,9,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm13, %xmm0, %xmm0
-; AVX512DQ-FCP-NEXT: vpshufb %ymm2, %ymm9, %ymm2
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm0 = [4,5,2,3,2,3,0,1,10,11,8,9,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm2, %xmm2
+; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm9, %ymm4
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm8[3,2,3,3,7,6,7,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2],ymm5[3,4],ymm2[5,6,7,8],ymm5[9],ymm2[10],ymm5[11,12],ymm2[13,14,15]
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm6 = [2,2,3,2,8,9,8,9]
-; AVX512DQ-FCP-NEXT: vpermi2q %zmm0, %zmm2, %zmm6
-; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %xmm0
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2],ymm5[3,4],ymm4[5,6,7,8],ymm5[9],ymm4[10],ymm5[11,12],ymm4[13,14,15]
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm8 = [2,2,3,2,8,9,8,9]
+; AVX512DQ-FCP-NEXT: vpermi2q %zmm2, %zmm4, %zmm8
+; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %xmm6
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %xmm2
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm22, %xmm5
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm22, %xmm4
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm5 = [0,1,2,3,8,9,10,11,4,5,4,5,6,7,12,13]
; AVX512DQ-FCP-NEXT: vpshufb %xmm5, %xmm2, %xmm2
; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm15, %ymm1
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm8 = ymm12[2,3,2,3,6,7,6,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7,8],ymm8[9],ymm1[10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm8 = [2,3,2,2,8,9,8,9]
-; AVX512DQ-FCP-NEXT: vpermi2q %zmm2, %zmm1, %zmm8
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm6 ^ (zmm17 & (zmm8 ^ zmm6))
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm12[2,3,2,3,6,7,6,7]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2],ymm4[3],ymm1[4,5],ymm4[6],ymm1[7,8],ymm4[9],ymm1[10],ymm4[11],ymm1[12,13],ymm4[14],ymm1[15]
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm4 = [2,3,2,2,8,9,8,9]
+; AVX512DQ-FCP-NEXT: vpermi2q %zmm2, %zmm1, %zmm4
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm8 ^ (zmm17 & (zmm4 ^ zmm8))
; AVX512DQ-FCP-NEXT: vpbroadcastq 24(%r8), %ymm1
; AVX512DQ-FCP-NEXT: vpbroadcastq 32(%r8), %ymm2
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (mem & (zmm1 ^ zmm8))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (mem & (zmm1 ^ zmm4))
; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %xmm2
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm21, %xmm6
-; AVX512DQ-FCP-NEXT: vpshufb %xmm6, %xmm2, %xmm6
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm21, %xmm4
+; AVX512DQ-FCP-NEXT: vpshufb %xmm4, %xmm2, %xmm4
; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %xmm8
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm11 = xmm8[1,2,2,2]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0],xmm11[1],xmm6[2],xmm11[3],xmm6[4,5],xmm11[6],xmm6[7]
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm9 = xmm8[1,2,2,2]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0],xmm9[1],xmm4[2],xmm9[3],xmm4[4,5],xmm9[6],xmm4[7]
; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm13, %xmm2, %xmm2
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm8 = [0,1,0,1,8,9,8,8]
-; AVX512DQ-FCP-NEXT: vpermi2q %zmm6, %zmm2, %zmm8
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm18, %xmm9
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm19, %xmm2
-; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm9, %xmm2
-; AVX512DQ-FCP-NEXT: vpbroadcastq 8(%rdi), %xmm6
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm6[1],xmm2[2,3],xmm6[4],xmm2[5],xmm6[6],xmm2[7]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm5, %xmm0, %xmm0
-; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,0,1],zmm2[0,1,0,1]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm8 ^ (zmm4 & (zmm0 ^ zmm8))
+; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm2, %xmm0
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm2 = [0,1,0,1,8,9,8,8]
+; AVX512DQ-FCP-NEXT: vpermi2q %zmm4, %zmm0, %zmm2
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm18, %xmm8
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm19, %xmm0
+; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm8, %xmm0
+; AVX512DQ-FCP-NEXT: vpbroadcastq 8(%rdi), %xmm4
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm4[1],xmm0[2,3],xmm4[4],xmm0[5],xmm4[6],xmm0[7]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm8[0],xmm6[1],xmm8[1],xmm6[2],xmm8[2],xmm6[3],xmm8[3]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm5, %xmm4, %xmm4
+; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm4[0,1,0,1],zmm0[0,1,0,1]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm2 ^ (zmm14 & (zmm0 ^ zmm2))
; AVX512DQ-FCP-NEXT: vpbroadcastq (%r8), %ymm2
; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm4 = ymm20[0,1,1,1]
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm2, %zmm2
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-6.ll
index ca8fcf2ee0f2c..5e26564465c25 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-6.ll
@@ -4138,213 +4138,211 @@ define void @store_i16_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX512-FCP-LABEL: store_i16_stride6_vf32:
; AVX512-FCP: # %bb.0:
-; AVX512-FCP-NEXT: subq $40, %rsp
+; AVX512-FCP-NEXT: subq $72, %rsp
; AVX512-FCP-NEXT: vmovdqa 32(%rcx), %ymm1
; AVX512-FCP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %ymm3
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm3[4],ymm1[4],ymm3[5],ymm1[5],ymm3[6],ymm1[6],ymm3[7],ymm1[7],ymm3[12],ymm1[12],ymm3[13],ymm1[13],ymm3[14],ymm1[14],ymm3[15],ymm1[15]
+; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %ymm2
+; AVX512-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm2[4],ymm1[4],ymm2[5],ymm1[5],ymm2[6],ymm1[6],ymm2[7],ymm1[7],ymm2[12],ymm1[12],ymm2[13],ymm1[13],ymm2[14],ymm1[14],ymm2[15],ymm1[15]
; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm1 = ymm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm1[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm3[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vmovdqa64 %ymm3, %ymm22
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm2[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm4 = [4,5,4,5,4,5,4,5,21,22,21,22,21,22,23,23]
; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm4, %zmm1
; AVX512-FCP-NEXT: vmovdqa 32(%rsi), %ymm5
+; AVX512-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm0 = [u,u,u,u,4,5,10,11,u,u,u,u,u,u,u,u,24,25,22,23,20,21,26,27,u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm5, %ymm2
; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %ymm6
+; AVX512-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm6, %ymm3
; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm6[4],ymm5[4],ymm6[5],ymm5[5],ymm6[6],ymm5[6],ymm6[7],ymm5[7],ymm6[12],ymm5[12],ymm6[13],ymm5[13],ymm6[14],ymm5[14],ymm6[15],ymm5[15]
-; AVX512-FCP-NEXT: vmovdqa64 %ymm6, %ymm23
-; AVX512-FCP-NEXT: vmovdqa64 %ymm5, %ymm25
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm5 = [2,1,2,3,11,11,11,11]
; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm5, %zmm2
; AVX512-FCP-NEXT: movw $18724, %ax # imm = 0x4924
; AVX512-FCP-NEXT: kmovw %eax, %k1
; AVX512-FCP-NEXT: vmovdqa32 %zmm1, %zmm2 {%k1}
-; AVX512-FCP-NEXT: vmovdqa 32(%r8), %ymm9
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm1 = ymm9[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm6 = [12,1,2,13,4,5,14,7]
+; AVX512-FCP-NEXT: vmovdqa 32(%r8), %ymm7
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm1 = ymm7[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm8 = [12,1,2,13,4,5,14,7]
; AVX512-FCP-NEXT: vmovdqa %ymm2, %ymm3
-; AVX512-FCP-NEXT: vpermt2d %ymm1, %ymm6, %ymm3
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm7 = [8,21,10,11,20,13,14,23]
+; AVX512-FCP-NEXT: vpermt2d %ymm1, %ymm8, %ymm3
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm9 = [8,21,10,11,20,13,14,23]
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm1 = [u,u,u,u,u,u,u,u,14,15,14,15,14,15,14,15,28,29,26,27,26,27,30,31,30,31,30,31,30,31,30,31]
-; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm9, %ymm8
-; AVX512-FCP-NEXT: vmovdqa64 %ymm9, %ymm24
-; AVX512-FCP-NEXT: vpermt2d %zmm8, %zmm7, %zmm2
+; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm7, %ymm6
+; AVX512-FCP-NEXT: vmovdqa64 %ymm7, %ymm30
+; AVX512-FCP-NEXT: vpermt2d %zmm6, %zmm9, %zmm2
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 32(%r9), %ymm9
+; AVX512-FCP-NEXT: vmovdqu64 %zmm2, (%rsp) # 64-byte Spill
+; AVX512-FCP-NEXT: vmovdqa 32(%r9), %ymm7
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm2 = [u,u,u,u,u,u,u,u,8,9,10,11,12,13,14,15,24,25,28,29,28,29,26,27,24,25,26,27,28,29,30,31]
-; AVX512-FCP-NEXT: vpshufb %ymm2, %ymm9, %ymm8
+; AVX512-FCP-NEXT: vpshufb %ymm2, %ymm7, %ymm6
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [16,17,20,21,20,21,22,23,24,25,24,25,24,25,24,25,16,17,20,21,20,21,22,23,24,25,24,25,24,25,24,25]
; AVX512-FCP-NEXT: # ymm3 = mem[0,1,0,1]
-; AVX512-FCP-NEXT: vpshufb %ymm3, %ymm9, %ymm10
-; AVX512-FCP-NEXT: vmovdqa64 %ymm9, %ymm27
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm9 = [2,2,2,3,10,9,10,11]
-; AVX512-FCP-NEXT: vpermt2q %zmm8, %zmm9, %zmm10
-; AVX512-FCP-NEXT: vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vmovdqa (%rcx), %ymm11
-; AVX512-FCP-NEXT: vmovdqa (%rdx), %ymm12
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm8 = ymm12[4],ymm11[4],ymm12[5],ymm11[5],ymm12[6],ymm11[6],ymm12[7],ymm11[7],ymm12[12],ymm11[12],ymm12[13],ymm11[13],ymm12[14],ymm11[14],ymm12[15],ymm11[15]
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm10 = ymm11[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm11[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vmovdqa64 %ymm11, %ymm28
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm11 = ymm12[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm12[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vmovdqa64 %ymm12, %ymm29
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm10 = ymm11[0],ymm10[0],ymm11[1],ymm10[1],ymm11[2],ymm10[2],ymm11[3],ymm10[3],ymm11[8],ymm10[8],ymm11[9],ymm10[9],ymm11[10],ymm10[10],ymm11[11],ymm10[11]
-; AVX512-FCP-NEXT: vpermt2d %zmm8, %zmm4, %zmm10
-; AVX512-FCP-NEXT: vmovdqa (%rsi), %ymm15
-; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm15, %ymm4
+; AVX512-FCP-NEXT: vpshufb %ymm3, %ymm7, %ymm10
+; AVX512-FCP-NEXT: vmovdqa64 %ymm7, %ymm31
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm19 = [2,2,0,3,10,0,10,11]
+; AVX512-FCP-NEXT: vpermi2q %zmm6, %zmm10, %zmm19
+; AVX512-FCP-NEXT: vmovdqa (%rcx), %ymm7
+; AVX512-FCP-NEXT: vmovdqa (%rdx), %ymm10
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm6 = ymm10[4],ymm7[4],ymm10[5],ymm7[5],ymm10[6],ymm7[6],ymm10[7],ymm7[7],ymm10[12],ymm7[12],ymm10[13],ymm7[13],ymm10[14],ymm7[14],ymm10[15],ymm7[15]
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm12 = ymm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm7[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vmovdqa64 %ymm7, %ymm28
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm13 = ymm10[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm10[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vmovdqa64 %ymm10, %ymm27
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm12 = ymm13[0],ymm12[0],ymm13[1],ymm12[1],ymm13[2],ymm12[2],ymm13[3],ymm12[3],ymm13[8],ymm12[8],ymm13[9],ymm12[9],ymm13[10],ymm12[10],ymm13[11],ymm12[11]
+; AVX512-FCP-NEXT: vpermt2d %zmm6, %zmm4, %zmm12
+; AVX512-FCP-NEXT: vmovdqa (%rsi), %ymm6
+; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm6, %ymm4
; AVX512-FCP-NEXT: vmovdqa (%rdi), %ymm14
; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm14, %ymm0
; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[1],ymm4[1],ymm0[2],ymm4[2],ymm0[3],ymm4[3],ymm0[8],ymm4[8],ymm0[9],ymm4[9],ymm0[10],ymm4[10],ymm0[11],ymm4[11]
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm14[4],ymm15[4],ymm14[5],ymm15[5],ymm14[6],ymm15[6],ymm14[7],ymm15[7],ymm14[12],ymm15[12],ymm14[13],ymm15[13],ymm14[14],ymm15[14],ymm14[15],ymm15[15]
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm14[4],ymm6[4],ymm14[5],ymm6[5],ymm14[6],ymm6[6],ymm14[7],ymm6[7],ymm14[12],ymm6[12],ymm14[13],ymm6[13],ymm14[14],ymm6[14],ymm14[15],ymm6[15]
+; AVX512-FCP-NEXT: vmovdqa64 %ymm6, %ymm17
; AVX512-FCP-NEXT: vpermt2q %zmm4, %zmm5, %zmm0
-; AVX512-FCP-NEXT: vmovdqa32 %zmm10, %zmm0 {%k1}
-; AVX512-FCP-NEXT: vmovdqa (%r8), %ymm5
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm4 = ymm5[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
-; AVX512-FCP-NEXT: vpermi2d %ymm4, %ymm0, %ymm6
-; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm5, %ymm1
-; AVX512-FCP-NEXT: vmovdqa64 %ymm5, %ymm30
-; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm7, %zmm0
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm6, %zmm21
-; AVX512-FCP-NEXT: vmovdqa (%r9), %ymm1
-; AVX512-FCP-NEXT: vpshufb %ymm2, %ymm1, %ymm0
-; AVX512-FCP-NEXT: vpshufb %ymm3, %ymm1, %ymm11
-; AVX512-FCP-NEXT: vmovdqa64 %ymm1, %ymm16
-; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm9, %zmm11
-; AVX512-FCP-NEXT: vmovdqa 32(%rcx), %xmm2
-; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %xmm4
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm0 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm0 = [1,0,2,2,1,0,2,2,16,17,16,17,16,17,16,17]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm4, %xmm31
-; AVX512-FCP-NEXT: vmovdqa64 %xmm2, %xmm18
-; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm0, %zmm3
-; AVX512-FCP-NEXT: vmovdqa 32(%rsi), %xmm13
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
-; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm13, %xmm1
-; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %xmm12
-; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm12, %xmm4
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm4 = [8,8,0,9,0,1,0,1]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm12[0],xmm13[0],xmm12[1],xmm13[1],xmm12[2],xmm13[2],xmm12[3],xmm13[3]
-; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm4, %zmm1
+; AVX512-FCP-NEXT: vmovdqa32 %zmm12, %zmm0 {%k1}
+; AVX512-FCP-NEXT: vmovdqa (%r8), %ymm13
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm4 = ymm13[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
+; AVX512-FCP-NEXT: vpermi2d %ymm4, %ymm0, %ymm8
+; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm13, %ymm1
+; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm9, %zmm0
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm8, %zmm29
+; AVX512-FCP-NEXT: vmovdqa (%r9), %ymm4
+; AVX512-FCP-NEXT: vpshufb %ymm2, %ymm4, %ymm0
+; AVX512-FCP-NEXT: vpshufb %ymm3, %ymm4, %ymm1
+; AVX512-FCP-NEXT: vmovdqa64 %ymm4, %ymm26
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm20 = [2,2,2,3,10,9,10,11]
+; AVX512-FCP-NEXT: vpermi2q %zmm0, %zmm1, %zmm20
+; AVX512-FCP-NEXT: vmovdqa 32(%r9), %xmm3
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
+; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm3, %xmm1
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm3[0,0,2,1,4,5,6,7]
+; AVX512-FCP-NEXT: vmovdqa64 %xmm3, %xmm16
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm23 = [0,0,0,0,8,8,0,9]
+; AVX512-FCP-NEXT: vpermi2q %zmm1, %zmm2, %zmm23
+; AVX512-FCP-NEXT: vmovdqa (%r9), %xmm11
+; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm11, %xmm0
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm11[0,0,2,1,4,5,6,7]
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm18 = [0,0,0,0,8,8,0,9]
+; AVX512-FCP-NEXT: vpermi2q %zmm0, %zmm1, %zmm18
+; AVX512-FCP-NEXT: vmovdqa 32(%rcx), %xmm10
+; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %xmm9
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm0 = xmm10[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm9[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm21 = [1,0,2,2,1,0,2,2,16,17,16,17,16,17,16,17]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3]
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm21, %zmm1
+; AVX512-FCP-NEXT: vmovdqa 32(%rsi), %xmm8
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm6 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
+; AVX512-FCP-NEXT: vpshufb %xmm6, %xmm8, %xmm0
+; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %xmm7
+; AVX512-FCP-NEXT: vpshufb %xmm6, %xmm7, %xmm15
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm15[4],xmm0[4],xmm15[5],xmm0[5],xmm15[6],xmm0[6],xmm15[7],xmm0[7]
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm22 = [8,8,0,9,0,1,0,1]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; AVX512-FCP-NEXT: vpermt2q %zmm2, %zmm22, %zmm0
; AVX512-FCP-NEXT: movw $9362, %ax # imm = 0x2492
; AVX512-FCP-NEXT: kmovw %eax, %k2
-; AVX512-FCP-NEXT: vmovdqa32 %zmm3, %zmm1 {%k2}
-; AVX512-FCP-NEXT: vmovdqa %ymm1, %ymm3
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm20 = [16,9,10,17,12,13,18,15]
-; AVX512-FCP-NEXT: vmovdqa 32(%r8), %xmm10
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm6 = xmm10[2,1,3,3,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2d %zmm6, %zmm20, %zmm1
-; AVX512-FCP-NEXT: vpmovzxwd {{.*#+}} xmm6 = xmm10[0],zero,xmm10[1],zero,xmm10[2],zero,xmm10[3],zero
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm19 = [0,1,8,3,4,9,6,7]
-; AVX512-FCP-NEXT: vpermt2d %ymm6, %ymm19, %ymm3
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm3, %zmm17
-; AVX512-FCP-NEXT: vmovdqa (%rcx), %xmm9
-; AVX512-FCP-NEXT: vmovdqa (%rdx), %xmm7
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm9[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm3 = xmm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm7[0],xmm9[0],xmm7[1],xmm9[1],xmm7[2],xmm9[2],xmm7[3],xmm9[3]
-; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm0, %zmm3
-; AVX512-FCP-NEXT: vmovdqa (%rsi), %xmm5
+; AVX512-FCP-NEXT: vmovdqa32 %zmm1, %zmm0 {%k2}
+; AVX512-FCP-NEXT: vmovdqa64 %ymm0, %ymm24
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm25 = [16,9,10,17,12,13,18,15]
+; AVX512-FCP-NEXT: vmovdqa 32(%r8), %xmm15
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm15[2,1,3,3,4,5,6,7]
+; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm25, %zmm0
+; AVX512-FCP-NEXT: vmovdqa (%rcx), %xmm5
+; AVX512-FCP-NEXT: vmovdqa (%rdx), %xmm4
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm12 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm12[0],xmm1[0],xmm12[1],xmm1[1],xmm12[2],xmm1[2],xmm12[3],xmm1[3]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm21, %zmm12
+; AVX512-FCP-NEXT: vmovdqa (%rsi), %xmm2
; AVX512-FCP-NEXT: vmovdqa (%rdi), %xmm1
-; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm5, %xmm8
-; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm1, %xmm2
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
-; AVX512-FCP-NEXT: vpermt2q %zmm8, %zmm4, %zmm2
-; AVX512-FCP-NEXT: vmovdqa32 %zmm3, %zmm2 {%k2}
-; AVX512-FCP-NEXT: vmovdqa (%r8), %xmm3
-; AVX512-FCP-NEXT: vpmovzxwd {{.*#+}} xmm4 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
-; AVX512-FCP-NEXT: vpermi2d %ymm4, %ymm2, %ymm19
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm3[2,1,3,3,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2d %zmm4, %zmm20, %zmm2
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm20 = [0,0,0,0,8,8,0,9]
-; AVX512-FCP-NEXT: vmovdqa 32(%r9), %xmm8
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
-; AVX512-FCP-NEXT: vpshufb %xmm4, %xmm8, %xmm0
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm6 = xmm8[0,0,2,1,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm20, %zmm6
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm19, %zmm19
-; AVX512-FCP-NEXT: vmovdqa (%r9), %xmm2
-; AVX512-FCP-NEXT: vpshufb %xmm4, %xmm2, %xmm0
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm2[0,0,2,1,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm20, %zmm4
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm14[0],ymm15[0],ymm14[1],ymm15[1],ymm14[2],ymm15[2],ymm14[3],ymm15[3],ymm14[8],ymm15[8],ymm14[9],ymm15[9],ymm14[10],ymm15[10],ymm14[11],ymm15[11]
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm26 = [1,1,1,1,10,10,10,11]
-; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm26, %zmm1
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm7[4],xmm9[4],xmm7[5],xmm9[5],xmm7[6],xmm9[6],xmm7[7],xmm9[7]
-; AVX512-FCP-NEXT: vmovdqa64 %ymm28, %ymm5
-; AVX512-FCP-NEXT: vmovdqa64 %ymm29, %ymm7
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm7 = ymm7[0],ymm5[0],ymm7[1],ymm5[1],ymm7[2],ymm5[2],ymm7[3],ymm5[3],ymm7[8],ymm5[8],ymm7[9],ymm5[9],ymm7[10],ymm5[10],ymm7[11],ymm5[11]
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm9 = [17,18,17,18,0,0,19,19,5,4,2,2,5,4,6,6]
-; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm9, %zmm7
-; AVX512-FCP-NEXT: vmovdqa32 %zmm1, %zmm7 {%k1}
+; AVX512-FCP-NEXT: vpshufb %xmm6, %xmm2, %xmm3
+; AVX512-FCP-NEXT: vpshufb %xmm6, %xmm1, %xmm6
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm6[4],xmm3[4],xmm6[5],xmm3[5],xmm6[6],xmm3[6],xmm6[7],xmm3[7]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm22, %zmm3
+; AVX512-FCP-NEXT: vpmovzxwd {{.*#+}} xmm6 = xmm15[0],zero,xmm15[1],zero,xmm15[2],zero,xmm15[3],zero
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm22 = [0,1,8,3,4,9,6,7]
+; AVX512-FCP-NEXT: vpermt2d %ymm6, %ymm22, %ymm24
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm24, %zmm21
+; AVX512-FCP-NEXT: vmovdqa32 %zmm12, %zmm3 {%k2}
+; AVX512-FCP-NEXT: vmovdqa (%r8), %xmm6
+; AVX512-FCP-NEXT: vpmovzxwd {{.*#+}} xmm0 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
+; AVX512-FCP-NEXT: vpermi2d %ymm0, %ymm3, %ymm22
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm6[2,1,3,3,4,5,6,7]
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm25, %zmm3
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm3, %zmm22, %zmm22
+; AVX512-FCP-NEXT: vmovdqa64 %ymm17, %ymm0
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm14[0],ymm0[0],ymm14[1],ymm0[1],ymm14[2],ymm0[2],ymm14[3],ymm0[3],ymm14[8],ymm0[8],ymm14[9],ymm0[9],ymm14[10],ymm0[10],ymm14[11],ymm0[11]
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm3 = [1,1,1,1,10,10,10,11]
+; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm3, %zmm1
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; AVX512-FCP-NEXT: vmovdqa64 %ymm28, %ymm2
+; AVX512-FCP-NEXT: vmovdqa64 %ymm27, %ymm4
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[8],ymm2[8],ymm4[9],ymm2[9],ymm4[10],ymm2[10],ymm4[11],ymm2[11]
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm4 = [17,18,17,18,0,0,19,19,5,4,2,2,5,4,6,6]
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm4, %zmm2
+; AVX512-FCP-NEXT: vmovdqa32 %zmm1, %zmm2 {%k1}
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm0 = [12,13,10,11,10,11,14,15,14,15,14,15,14,15,14,15]
-; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm3, %xmm1
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm3 = [0,9,2,3,8,5,6,11]
-; AVX512-FCP-NEXT: vmovdqa %ymm7, %ymm14
-; AVX512-FCP-NEXT: vpermt2d %ymm1, %ymm3, %ymm14
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm15 = [8,9,20,11,12,21,14,15]
-; AVX512-FCP-NEXT: vmovdqa64 %ymm30, %ymm1
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm1 = ymm1[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm15, %zmm7
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm7, %zmm14, %zmm7
-; AVX512-FCP-NEXT: vmovdqa64 %ymm16, %ymm1
+; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm6, %xmm1
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm5 = [0,9,2,3,8,5,6,11]
+; AVX512-FCP-NEXT: vmovdqa %ymm2, %ymm6
+; AVX512-FCP-NEXT: vpermt2d %ymm1, %ymm5, %ymm6
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm12 = [8,9,20,11,12,21,14,15]
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm1 = ymm13[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
+; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm12, %zmm2
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm6, %zmm6
+; AVX512-FCP-NEXT: vmovdqa64 %ymm26, %ymm1
; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm14 = ymm1[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm1 = [8,9,12,13,12,13,10,11,8,9,10,11,12,13,14,15]
-; AVX512-FCP-NEXT: vpshufb %xmm1, %xmm2, %xmm2
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm20 = [0,1,0,1,10,10,10,10]
-; AVX512-FCP-NEXT: vpermt2q %zmm14, %zmm20, %zmm2
-; AVX512-FCP-NEXT: vmovdqa64 %ymm23, %ymm14
-; AVX512-FCP-NEXT: vmovdqa64 %ymm25, %ymm5
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm14 = ymm14[0],ymm5[0],ymm14[1],ymm5[1],ymm14[2],ymm5[2],ymm14[3],ymm5[3],ymm14[8],ymm5[8],ymm14[9],ymm5[9],ymm14[10],ymm5[10],ymm14[11],ymm5[11]
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm12 = xmm12[4],xmm13[4],xmm12[5],xmm13[5],xmm12[6],xmm13[6],xmm12[7],xmm13[7]
-; AVX512-FCP-NEXT: vpermt2q %zmm14, %zmm26, %zmm12
-; AVX512-FCP-NEXT: vmovdqa64 %xmm31, %xmm5
-; AVX512-FCP-NEXT: vmovdqa64 %xmm18, %xmm13
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm5[4],xmm13[4],xmm5[5],xmm13[5],xmm5[6],xmm13[6],xmm5[7],xmm13[7]
-; AVX512-FCP-NEXT: vmovdqa64 %ymm22, %ymm14
-; AVX512-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm13 # 32-byte Folded Reload
-; AVX512-FCP-NEXT: # ymm13 = ymm14[0],mem[0],ymm14[1],mem[1],ymm14[2],mem[2],ymm14[3],mem[3],ymm14[8],mem[8],ymm14[9],mem[9],ymm14[10],mem[10],ymm14[11],mem[11]
-; AVX512-FCP-NEXT: vpermt2d %zmm5, %zmm9, %zmm13
-; AVX512-FCP-NEXT: vmovdqa32 %zmm12, %zmm13 {%k1}
-; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm10, %xmm0
-; AVX512-FCP-NEXT: vpermi2d %ymm0, %ymm13, %ymm3
-; AVX512-FCP-NEXT: vmovdqa64 %ymm24, %ymm0
+; AVX512-FCP-NEXT: vpshufb %xmm1, %xmm11, %xmm2
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm11 = [0,0,0,1,0,10,10,0]
+; AVX512-FCP-NEXT: vpermt2q %zmm14, %zmm11, %zmm2
+; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX512-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: # ymm14 = ymm14[0],mem[0],ymm14[1],mem[1],ymm14[2],mem[2],ymm14[3],mem[3],ymm14[8],mem[8],ymm14[9],mem[9],ymm14[10],mem[10],ymm14[11],mem[11]
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm7 = xmm7[4],xmm8[4],xmm7[5],xmm8[5],xmm7[6],xmm8[6],xmm7[7],xmm8[7]
+; AVX512-FCP-NEXT: vpermt2q %zmm14, %zmm3, %zmm7
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
+; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX512-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: # ymm8 = ymm8[0],mem[0],ymm8[1],mem[1],ymm8[2],mem[2],ymm8[3],mem[3],ymm8[8],mem[8],ymm8[9],mem[9],ymm8[10],mem[10],ymm8[11],mem[11]
+; AVX512-FCP-NEXT: vpermt2d %zmm3, %zmm4, %zmm8
+; AVX512-FCP-NEXT: vmovdqa32 %zmm7, %zmm8 {%k1}
+; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm15, %xmm0
+; AVX512-FCP-NEXT: vpermi2d %ymm0, %ymm8, %ymm5
+; AVX512-FCP-NEXT: vmovdqa64 %ymm30, %ymm0
; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm15, %zmm13
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm13, %zmm3, %zmm0
-; AVX512-FCP-NEXT: vpshufb %xmm1, %xmm8, %xmm1
-; AVX512-FCP-NEXT: vmovdqa64 %ymm27, %ymm3
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm12, %zmm8
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm5, %zmm0
+; AVX512-FCP-NEXT: vmovdqa64 %xmm16, %xmm3
+; AVX512-FCP-NEXT: vpshufb %xmm1, %xmm3, %xmm1
+; AVX512-FCP-NEXT: vmovdqa64 %ymm31, %ymm3
; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm3 = ymm3[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
-; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm20, %zmm1
+; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm11, %zmm1
; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (zmm3 & (zmm1 ^ zmm0))
; AVX512-FCP-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512-FCP-NEXT: vmovdqa64 %zmm1, 256(%rax)
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm3 & (zmm2 ^ zmm7))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm3 & (zmm2 ^ zmm6))
; AVX512-FCP-NEXT: vmovdqa64 %zmm2, 64(%rax)
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm4 ^ (zmm0 & (zmm4 ^ zmm19))
-; AVX512-FCP-NEXT: vmovdqa64 %zmm4, (%rax)
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm6 ^ (zmm0 & (zmm6 ^ zmm17))
-; AVX512-FCP-NEXT: vmovdqa64 %zmm6, 192(%rax)
; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm11 = zmm11 ^ (zmm0 & (zmm11 ^ zmm21))
-; AVX512-FCP-NEXT: vmovdqa64 %zmm11, 128(%rax)
-; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
-; AVX512-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 # 64-byte Folded Reload
-; AVX512-FCP-NEXT: # zmm1 = zmm1 ^ (zmm0 & (zmm1 ^ mem))
-; AVX512-FCP-NEXT: vmovdqa64 %zmm1, 320(%rax)
-; AVX512-FCP-NEXT: addq $40, %rsp
+; AVX512-FCP-NEXT: vpternlogq $184, (%rsp), %zmm0, %zmm19 # 64-byte Folded Reload
+; AVX512-FCP-NEXT: # zmm19 = zmm19 ^ (zmm0 & (zmm19 ^ mem))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm20 = zmm20 ^ (zmm0 & (zmm20 ^ zmm29))
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm23 = zmm23 ^ (zmm0 & (zmm23 ^ zmm21))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm18 = zmm18 ^ (zmm0 & (zmm18 ^ zmm22))
+; AVX512-FCP-NEXT: vmovdqa64 %zmm18, (%rax)
+; AVX512-FCP-NEXT: vmovdqa64 %zmm23, 192(%rax)
+; AVX512-FCP-NEXT: vmovdqa64 %zmm20, 128(%rax)
+; AVX512-FCP-NEXT: vmovdqa64 %zmm19, 320(%rax)
+; AVX512-FCP-NEXT: addq $72, %rsp
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
@@ -4574,213 +4572,211 @@ define void @store_i16_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX512DQ-FCP-LABEL: store_i16_stride6_vf32:
; AVX512DQ-FCP: # %bb.0:
-; AVX512DQ-FCP-NEXT: subq $40, %rsp
+; AVX512DQ-FCP-NEXT: subq $72, %rsp
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rcx), %ymm1
; AVX512DQ-FCP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %ymm3
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm3[4],ymm1[4],ymm3[5],ymm1[5],ymm3[6],ymm1[6],ymm3[7],ymm1[7],ymm3[12],ymm1[12],ymm3[13],ymm1[13],ymm3[14],ymm1[14],ymm3[15],ymm1[15]
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %ymm2
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm2[4],ymm1[4],ymm2[5],ymm1[5],ymm2[6],ymm1[6],ymm2[7],ymm1[7],ymm2[12],ymm1[12],ymm2[13],ymm1[13],ymm2[14],ymm1[14],ymm2[15],ymm1[15]
; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm1 = ymm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm1[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm3[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm3, %ymm22
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm2[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm4 = [4,5,4,5,4,5,4,5,21,22,21,22,21,22,23,23]
; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm4, %zmm1
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rsi), %ymm5
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm0 = [u,u,u,u,4,5,10,11,u,u,u,u,u,u,u,u,24,25,22,23,20,21,26,27,u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm5, %ymm2
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %ymm6
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm6, %ymm3
; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm6[4],ymm5[4],ymm6[5],ymm5[5],ymm6[6],ymm5[6],ymm6[7],ymm5[7],ymm6[12],ymm5[12],ymm6[13],ymm5[13],ymm6[14],ymm5[14],ymm6[15],ymm5[15]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm6, %ymm23
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm5, %ymm25
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm5 = [2,1,2,3,11,11,11,11]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm5, %zmm2
; AVX512DQ-FCP-NEXT: movw $18724, %ax # imm = 0x4924
; AVX512DQ-FCP-NEXT: kmovw %eax, %k1
; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm1, %zmm2 {%k1}
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %ymm9
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm1 = ymm9[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm6 = [12,1,2,13,4,5,14,7]
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %ymm7
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm1 = ymm7[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm8 = [12,1,2,13,4,5,14,7]
; AVX512DQ-FCP-NEXT: vmovdqa %ymm2, %ymm3
-; AVX512DQ-FCP-NEXT: vpermt2d %ymm1, %ymm6, %ymm3
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm7 = [8,21,10,11,20,13,14,23]
+; AVX512DQ-FCP-NEXT: vpermt2d %ymm1, %ymm8, %ymm3
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm9 = [8,21,10,11,20,13,14,23]
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm1 = [u,u,u,u,u,u,u,u,14,15,14,15,14,15,14,15,28,29,26,27,26,27,30,31,30,31,30,31,30,31,30,31]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm9, %ymm8
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm9, %ymm24
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm8, %zmm7, %zmm2
+; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm7, %ymm6
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm7, %ymm30
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm6, %zmm9, %zmm2
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%r9), %ymm9
+; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm2, (%rsp) # 64-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%r9), %ymm7
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm2 = [u,u,u,u,u,u,u,u,8,9,10,11,12,13,14,15,24,25,28,29,28,29,26,27,24,25,26,27,28,29,30,31]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm2, %ymm9, %ymm8
+; AVX512DQ-FCP-NEXT: vpshufb %ymm2, %ymm7, %ymm6
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [16,17,20,21,20,21,22,23,24,25,24,25,24,25,24,25,16,17,20,21,20,21,22,23,24,25,24,25,24,25,24,25]
; AVX512DQ-FCP-NEXT: # ymm3 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm3, %ymm9, %ymm10
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm9, %ymm27
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm9 = [2,2,2,3,10,9,10,11]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm8, %zmm9, %zmm10
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %ymm11
-; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %ymm12
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm8 = ymm12[4],ymm11[4],ymm12[5],ymm11[5],ymm12[6],ymm11[6],ymm12[7],ymm11[7],ymm12[12],ymm11[12],ymm12[13],ymm11[13],ymm12[14],ymm11[14],ymm12[15],ymm11[15]
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm10 = ymm11[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm11[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm11, %ymm28
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm11 = ymm12[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm12[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm12, %ymm29
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm10 = ymm11[0],ymm10[0],ymm11[1],ymm10[1],ymm11[2],ymm10[2],ymm11[3],ymm10[3],ymm11[8],ymm10[8],ymm11[9],ymm10[9],ymm11[10],ymm10[10],ymm11[11],ymm10[11]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm8, %zmm4, %zmm10
-; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %ymm15
-; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm15, %ymm4
+; AVX512DQ-FCP-NEXT: vpshufb %ymm3, %ymm7, %ymm10
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm7, %ymm31
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm19 = [2,2,0,3,10,0,10,11]
+; AVX512DQ-FCP-NEXT: vpermi2q %zmm6, %zmm10, %zmm19
+; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %ymm7
+; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %ymm10
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm6 = ymm10[4],ymm7[4],ymm10[5],ymm7[5],ymm10[6],ymm7[6],ymm10[7],ymm7[7],ymm10[12],ymm7[12],ymm10[13],ymm7[13],ymm10[14],ymm7[14],ymm10[15],ymm7[15]
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm12 = ymm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm7[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm7, %ymm28
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm13 = ymm10[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm10[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm10, %ymm27
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm12 = ymm13[0],ymm12[0],ymm13[1],ymm12[1],ymm13[2],ymm12[2],ymm13[3],ymm12[3],ymm13[8],ymm12[8],ymm13[9],ymm12[9],ymm13[10],ymm12[10],ymm13[11],ymm12[11]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm6, %zmm4, %zmm12
+; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %ymm6
+; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm6, %ymm4
; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %ymm14
; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm14, %ymm0
; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[1],ymm4[1],ymm0[2],ymm4[2],ymm0[3],ymm4[3],ymm0[8],ymm4[8],ymm0[9],ymm4[9],ymm0[10],ymm4[10],ymm0[11],ymm4[11]
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm14[4],ymm15[4],ymm14[5],ymm15[5],ymm14[6],ymm15[6],ymm14[7],ymm15[7],ymm14[12],ymm15[12],ymm14[13],ymm15[13],ymm14[14],ymm15[14],ymm14[15],ymm15[15]
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm14[4],ymm6[4],ymm14[5],ymm6[5],ymm14[6],ymm6[6],ymm14[7],ymm6[7],ymm14[12],ymm6[12],ymm14[13],ymm6[13],ymm14[14],ymm6[14],ymm14[15],ymm6[15]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm6, %ymm17
; AVX512DQ-FCP-NEXT: vpermt2q %zmm4, %zmm5, %zmm0
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm10, %zmm0 {%k1}
-; AVX512DQ-FCP-NEXT: vmovdqa (%r8), %ymm5
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm4 = ymm5[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpermi2d %ymm4, %ymm0, %ymm6
-; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm5, %ymm1
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm5, %ymm30
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm7, %zmm0
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm6, %zmm21
-; AVX512DQ-FCP-NEXT: vmovdqa (%r9), %ymm1
-; AVX512DQ-FCP-NEXT: vpshufb %ymm2, %ymm1, %ymm0
-; AVX512DQ-FCP-NEXT: vpshufb %ymm3, %ymm1, %ymm11
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm1, %ymm16
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm9, %zmm11
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rcx), %xmm2
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %xmm4
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm0 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm0 = [1,0,2,2,1,0,2,2,16,17,16,17,16,17,16,17]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm4, %xmm31
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm2, %xmm18
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm0, %zmm3
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rsi), %xmm13
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm13, %xmm1
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %xmm12
-; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm12, %xmm4
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm4 = [8,8,0,9,0,1,0,1]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm12[0],xmm13[0],xmm12[1],xmm13[1],xmm12[2],xmm13[2],xmm12[3],xmm13[3]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm4, %zmm1
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm12, %zmm0 {%k1}
+; AVX512DQ-FCP-NEXT: vmovdqa (%r8), %ymm13
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm4 = ymm13[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
+; AVX512DQ-FCP-NEXT: vpermi2d %ymm4, %ymm0, %ymm8
+; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm13, %ymm1
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm9, %zmm0
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm8, %zmm29
+; AVX512DQ-FCP-NEXT: vmovdqa (%r9), %ymm4
+; AVX512DQ-FCP-NEXT: vpshufb %ymm2, %ymm4, %ymm0
+; AVX512DQ-FCP-NEXT: vpshufb %ymm3, %ymm4, %ymm1
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm4, %ymm26
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm20 = [2,2,2,3,10,9,10,11]
+; AVX512DQ-FCP-NEXT: vpermi2q %zmm0, %zmm1, %zmm20
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%r9), %xmm3
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm3, %xmm1
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm3[0,0,2,1,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm3, %xmm16
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm23 = [0,0,0,0,8,8,0,9]
+; AVX512DQ-FCP-NEXT: vpermi2q %zmm1, %zmm2, %zmm23
+; AVX512DQ-FCP-NEXT: vmovdqa (%r9), %xmm11
+; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm11, %xmm0
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm11[0,0,2,1,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm18 = [0,0,0,0,8,8,0,9]
+; AVX512DQ-FCP-NEXT: vpermi2q %zmm0, %zmm1, %zmm18
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rcx), %xmm10
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %xmm9
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm0 = xmm10[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm9[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm21 = [1,0,2,2,1,0,2,2,16,17,16,17,16,17,16,17]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm21, %zmm1
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rsi), %xmm8
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm6 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm6, %xmm8, %xmm0
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %xmm7
+; AVX512DQ-FCP-NEXT: vpshufb %xmm6, %xmm7, %xmm15
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm15[4],xmm0[4],xmm15[5],xmm0[5],xmm15[6],xmm0[6],xmm15[7],xmm0[7]
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm22 = [8,8,0,9,0,1,0,1]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm2, %zmm22, %zmm0
; AVX512DQ-FCP-NEXT: movw $9362, %ax # imm = 0x2492
; AVX512DQ-FCP-NEXT: kmovw %eax, %k2
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm3, %zmm1 {%k2}
-; AVX512DQ-FCP-NEXT: vmovdqa %ymm1, %ymm3
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm20 = [16,9,10,17,12,13,18,15]
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %xmm10
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm6 = xmm10[2,1,3,3,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm6, %zmm20, %zmm1
-; AVX512DQ-FCP-NEXT: vpmovzxwd {{.*#+}} xmm6 = xmm10[0],zero,xmm10[1],zero,xmm10[2],zero,xmm10[3],zero
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm19 = [0,1,8,3,4,9,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2d %ymm6, %ymm19, %ymm3
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm3, %zmm17
-; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %xmm9
-; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %xmm7
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm9[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm3 = xmm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm7[0],xmm9[0],xmm7[1],xmm9[1],xmm7[2],xmm9[2],xmm7[3],xmm9[3]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm0, %zmm3
-; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %xmm5
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm1, %zmm0 {%k2}
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm0, %ymm24
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm25 = [16,9,10,17,12,13,18,15]
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %xmm15
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm1 = xmm15[2,1,3,3,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm25, %zmm0
+; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %xmm5
+; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %xmm4
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm12 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm12[0],xmm1[0],xmm12[1],xmm1[1],xmm12[2],xmm1[2],xmm12[3],xmm1[3]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm21, %zmm12
+; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %xmm2
; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %xmm1
-; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm5, %xmm8
-; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm1, %xmm2
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm8, %zmm4, %zmm2
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm3, %zmm2 {%k2}
-; AVX512DQ-FCP-NEXT: vmovdqa (%r8), %xmm3
-; AVX512DQ-FCP-NEXT: vpmovzxwd {{.*#+}} xmm4 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
-; AVX512DQ-FCP-NEXT: vpermi2d %ymm4, %ymm2, %ymm19
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm3[2,1,3,3,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm4, %zmm20, %zmm2
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm20 = [0,0,0,0,8,8,0,9]
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%r9), %xmm8
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm4, %xmm8, %xmm0
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm6 = xmm8[0,0,2,1,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm20, %zmm6
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm19, %zmm19
-; AVX512DQ-FCP-NEXT: vmovdqa (%r9), %xmm2
-; AVX512DQ-FCP-NEXT: vpshufb %xmm4, %xmm2, %xmm0
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm2[0,0,2,1,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm20, %zmm4
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm14[0],ymm15[0],ymm14[1],ymm15[1],ymm14[2],ymm15[2],ymm14[3],ymm15[3],ymm14[8],ymm15[8],ymm14[9],ymm15[9],ymm14[10],ymm15[10],ymm14[11],ymm15[11]
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm26 = [1,1,1,1,10,10,10,11]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm26, %zmm1
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm7[4],xmm9[4],xmm7[5],xmm9[5],xmm7[6],xmm9[6],xmm7[7],xmm9[7]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm28, %ymm5
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm29, %ymm7
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm7 = ymm7[0],ymm5[0],ymm7[1],ymm5[1],ymm7[2],ymm5[2],ymm7[3],ymm5[3],ymm7[8],ymm5[8],ymm7[9],ymm5[9],ymm7[10],ymm5[10],ymm7[11],ymm5[11]
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm9 = [17,18,17,18,0,0,19,19,5,4,2,2,5,4,6,6]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm9, %zmm7
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm1, %zmm7 {%k1}
+; AVX512DQ-FCP-NEXT: vpshufb %xmm6, %xmm2, %xmm3
+; AVX512DQ-FCP-NEXT: vpshufb %xmm6, %xmm1, %xmm6
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm6[4],xmm3[4],xmm6[5],xmm3[5],xmm6[6],xmm3[6],xmm6[7],xmm3[7]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm22, %zmm3
+; AVX512DQ-FCP-NEXT: vpmovzxwd {{.*#+}} xmm6 = xmm15[0],zero,xmm15[1],zero,xmm15[2],zero,xmm15[3],zero
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm22 = [0,1,8,3,4,9,6,7]
+; AVX512DQ-FCP-NEXT: vpermt2d %ymm6, %ymm22, %ymm24
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm24, %zmm21
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm12, %zmm3 {%k2}
+; AVX512DQ-FCP-NEXT: vmovdqa (%r8), %xmm6
+; AVX512DQ-FCP-NEXT: vpmovzxwd {{.*#+}} xmm0 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
+; AVX512DQ-FCP-NEXT: vpermi2d %ymm0, %ymm3, %ymm22
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm0 = xmm6[2,1,3,3,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm25, %zmm3
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm3, %zmm22, %zmm22
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm17, %ymm0
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm14[0],ymm0[0],ymm14[1],ymm0[1],ymm14[2],ymm0[2],ymm14[3],ymm0[3],ymm14[8],ymm0[8],ymm14[9],ymm0[9],ymm14[10],ymm0[10],ymm14[11],ymm0[11]
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm3 = [1,1,1,1,10,10,10,11]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm3, %zmm1
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm28, %ymm2
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm27, %ymm4
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[8],ymm2[8],ymm4[9],ymm2[9],ymm4[10],ymm2[10],ymm4[11],ymm2[11]
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm4 = [17,18,17,18,0,0,19,19,5,4,2,2,5,4,6,6]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm4, %zmm2
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm1, %zmm2 {%k1}
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm0 = [12,13,10,11,10,11,14,15,14,15,14,15,14,15,14,15]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm3, %xmm1
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm3 = [0,9,2,3,8,5,6,11]
-; AVX512DQ-FCP-NEXT: vmovdqa %ymm7, %ymm14
-; AVX512DQ-FCP-NEXT: vpermt2d %ymm1, %ymm3, %ymm14
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm15 = [8,9,20,11,12,21,14,15]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm30, %ymm1
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm1 = ymm1[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm15, %zmm7
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm7, %zmm14, %zmm7
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm16, %ymm1
+; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm6, %xmm1
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm5 = [0,9,2,3,8,5,6,11]
+; AVX512DQ-FCP-NEXT: vmovdqa %ymm2, %ymm6
+; AVX512DQ-FCP-NEXT: vpermt2d %ymm1, %ymm5, %ymm6
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm12 = [8,9,20,11,12,21,14,15]
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm1 = ymm13[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm12, %zmm2
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm6, %zmm6
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm26, %ymm1
; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm14 = ymm1[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm1 = [8,9,12,13,12,13,10,11,8,9,10,11,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm1, %xmm2, %xmm2
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm20 = [0,1,0,1,10,10,10,10]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm14, %zmm20, %zmm2
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm23, %ymm14
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm25, %ymm5
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm14 = ymm14[0],ymm5[0],ymm14[1],ymm5[1],ymm14[2],ymm5[2],ymm14[3],ymm5[3],ymm14[8],ymm5[8],ymm14[9],ymm5[9],ymm14[10],ymm5[10],ymm14[11],ymm5[11]
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm12 = xmm12[4],xmm13[4],xmm12[5],xmm13[5],xmm12[6],xmm13[6],xmm12[7],xmm13[7]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm14, %zmm26, %zmm12
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm31, %xmm5
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm18, %xmm13
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm5[4],xmm13[4],xmm5[5],xmm13[5],xmm5[6],xmm13[6],xmm5[7],xmm13[7]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm22, %ymm14
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm13 # 32-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # ymm13 = ymm14[0],mem[0],ymm14[1],mem[1],ymm14[2],mem[2],ymm14[3],mem[3],ymm14[8],mem[8],ymm14[9],mem[9],ymm14[10],mem[10],ymm14[11],mem[11]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm5, %zmm9, %zmm13
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm12, %zmm13 {%k1}
-; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm10, %xmm0
-; AVX512DQ-FCP-NEXT: vpermi2d %ymm0, %ymm13, %ymm3
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm24, %ymm0
+; AVX512DQ-FCP-NEXT: vpshufb %xmm1, %xmm11, %xmm2
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm11 = [0,0,0,1,0,10,10,0]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm14, %zmm11, %zmm2
+; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # ymm14 = ymm14[0],mem[0],ymm14[1],mem[1],ymm14[2],mem[2],ymm14[3],mem[3],ymm14[8],mem[8],ymm14[9],mem[9],ymm14[10],mem[10],ymm14[11],mem[11]
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm7 = xmm7[4],xmm8[4],xmm7[5],xmm8[5],xmm7[6],xmm8[6],xmm7[7],xmm8[7]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm14, %zmm3, %zmm7
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
+; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # ymm8 = ymm8[0],mem[0],ymm8[1],mem[1],ymm8[2],mem[2],ymm8[3],mem[3],ymm8[8],mem[8],ymm8[9],mem[9],ymm8[10],mem[10],ymm8[11],mem[11]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm3, %zmm4, %zmm8
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm7, %zmm8 {%k1}
+; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm15, %xmm0
+; AVX512DQ-FCP-NEXT: vpermi2d %ymm0, %ymm8, %ymm5
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm30, %ymm0
; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm15, %zmm13
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm13, %zmm3, %zmm0
-; AVX512DQ-FCP-NEXT: vpshufb %xmm1, %xmm8, %xmm1
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm27, %ymm3
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm12, %zmm8
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm5, %zmm0
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm16, %xmm3
+; AVX512DQ-FCP-NEXT: vpshufb %xmm1, %xmm3, %xmm1
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm31, %ymm3
; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm3 = ymm3[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm20, %zmm1
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm11, %zmm1
; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (zmm3 & (zmm1 ^ zmm0))
; AVX512DQ-FCP-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm1, 256(%rax)
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm3 & (zmm2 ^ zmm7))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm3 & (zmm2 ^ zmm6))
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, 64(%rax)
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm4 ^ (zmm0 & (zmm4 ^ zmm19))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm4, (%rax)
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm6 ^ (zmm0 & (zmm6 ^ zmm17))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm6, 192(%rax)
; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm11 = zmm11 ^ (zmm0 & (zmm11 ^ zmm21))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm11, 128(%rax)
-; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
-; AVX512DQ-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 # 64-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # zmm1 = zmm1 ^ (zmm0 & (zmm1 ^ mem))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm1, 320(%rax)
-; AVX512DQ-FCP-NEXT: addq $40, %rsp
+; AVX512DQ-FCP-NEXT: vpternlogq $184, (%rsp), %zmm0, %zmm19 # 64-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # zmm19 = zmm19 ^ (zmm0 & (zmm19 ^ mem))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm20 = zmm20 ^ (zmm0 & (zmm20 ^ zmm29))
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm23 = zmm23 ^ (zmm0 & (zmm23 ^ zmm21))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm18 = zmm18 ^ (zmm0 & (zmm18 ^ zmm22))
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm18, (%rax)
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm23, 192(%rax)
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm20, 128(%rax)
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm19, 320(%rax)
+; AVX512DQ-FCP-NEXT: addq $72, %rsp
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
@@ -8749,7 +8745,7 @@ define void @store_i16_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX512-FCP-LABEL: store_i16_stride6_vf64:
; AVX512-FCP: # %bb.0:
-; AVX512-FCP-NEXT: subq $1320, %rsp # imm = 0x528
+; AVX512-FCP-NEXT: subq $1224, %rsp # imm = 0x4C8
; AVX512-FCP-NEXT: vmovdqa 96(%rcx), %ymm1
; AVX512-FCP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vmovdqa 96(%rdx), %ymm2
@@ -8757,324 +8753,324 @@ define void @store_i16_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm2[4],ymm1[4],ymm2[5],ymm1[5],ymm2[6],ymm1[6],ymm2[7],ymm1[7],ymm2[12],ymm1[12],ymm2[13],ymm1[13],ymm2[14],ymm1[14],ymm2[15],ymm1[15]
; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm1 = ymm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm1[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm2[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm22 = [4,5,4,5,4,5,4,5,21,22,21,22,21,22,23,23]
-; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm22, %zmm1
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm22, %zmm2
; AVX512-FCP-NEXT: vmovdqa 96(%rsi), %ymm4
; AVX512-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm0 = [u,u,u,u,4,5,10,11,u,u,u,u,u,u,u,u,24,25,22,23,20,21,26,27,u,u,u,u,u,u,u,u]
-; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm4, %ymm2
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm1 = [u,u,u,u,4,5,10,11,u,u,u,u,u,u,u,u,24,25,22,23,20,21,26,27,u,u,u,u,u,u,u,u]
+; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm4, %ymm0
; AVX512-FCP-NEXT: vmovdqa 96(%rdi), %ymm5
; AVX512-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm5, %ymm3
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
+; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm5, %ymm3
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[1],ymm0[1],ymm3[2],ymm0[2],ymm3[3],ymm0[3],ymm3[8],ymm0[8],ymm3[9],ymm0[9],ymm3[10],ymm0[10],ymm3[11],ymm0[11]
; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm5[4],ymm4[4],ymm5[5],ymm4[5],ymm5[6],ymm4[6],ymm5[7],ymm4[7],ymm5[12],ymm4[12],ymm5[13],ymm4[13],ymm5[14],ymm4[14],ymm5[15],ymm4[15]
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm23 = [2,1,2,3,11,11,11,11]
-; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm23, %zmm2
+; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm23, %zmm0
; AVX512-FCP-NEXT: movw $18724, %ax # imm = 0x4924
; AVX512-FCP-NEXT: kmovw %eax, %k1
-; AVX512-FCP-NEXT: vmovdqa32 %zmm1, %zmm2 {%k1}
+; AVX512-FCP-NEXT: vmovdqa32 %zmm2, %zmm0 {%k1}
; AVX512-FCP-NEXT: vmovdqa 96(%r8), %ymm4
; AVX512-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm1 = ymm4[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm2 = ymm4[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm21 = [12,1,2,13,4,5,14,7]
-; AVX512-FCP-NEXT: vmovdqa %ymm2, %ymm3
-; AVX512-FCP-NEXT: vpermt2d %ymm1, %ymm21, %ymm3
+; AVX512-FCP-NEXT: vmovdqa %ymm0, %ymm3
+; AVX512-FCP-NEXT: vpermt2d %ymm2, %ymm21, %ymm3
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm24 = [8,21,10,11,20,13,14,23]
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm6 = [u,u,u,u,u,u,u,u,14,15,14,15,14,15,14,15,28,29,26,27,26,27,30,31,30,31,30,31,30,31,30,31]
-; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm4, %ymm1
-; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm24, %zmm2
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm1
-; AVX512-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm8 = [u,u,u,u,u,u,u,u,14,15,14,15,14,15,14,15,28,29,26,27,26,27,30,31,30,31,30,31,30,31,30,31]
+; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm4, %ymm2
+; AVX512-FCP-NEXT: vpermt2d %zmm2, %zmm24, %zmm0
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm3, %zmm0
+; AVX512-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT: vmovdqa 96(%r9), %ymm2
; AVX512-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm7 = [u,u,u,u,u,u,u,u,8,9,10,11,12,13,14,15,24,25,28,29,28,29,26,27,24,25,26,27,28,29,30,31]
-; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm2, %ymm1
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm9 = [u,u,u,u,u,u,u,u,8,9,10,11,12,13,14,15,24,25,28,29,28,29,26,27,24,25,26,27,28,29,30,31]
+; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm2, %ymm0
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm15 = [16,17,20,21,20,21,22,23,24,25,24,25,24,25,24,25,16,17,20,21,20,21,22,23,24,25,24,25,24,25,24,25]
; AVX512-FCP-NEXT: # ymm15 = mem[0,1,0,1]
; AVX512-FCP-NEXT: vpshufb %ymm15, %ymm2, %ymm2
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm25 = [2,2,2,3,10,9,10,11]
-; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm25, %zmm2
-; AVX512-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 64(%rcx), %ymm2
-; AVX512-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 64(%rdx), %ymm3
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm20 = [2,2,0,3,10,0,10,11]
+; AVX512-FCP-NEXT: vpermi2q %zmm0, %zmm2, %zmm20
+; AVX512-FCP-NEXT: vmovdqa 64(%rcx), %ymm3
; AVX512-FCP-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm1 = ymm3[4],ymm2[4],ymm3[5],ymm2[5],ymm3[6],ymm2[6],ymm3[7],ymm2[7],ymm3[12],ymm2[12],ymm3[13],ymm2[13],ymm3[14],ymm2[14],ymm3[15],ymm2[15]
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm2[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm3 = ymm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm3[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
-; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm22, %zmm2
+; AVX512-FCP-NEXT: vmovdqa 64(%rdx), %ymm4
+; AVX512-FCP-NEXT: vmovdqa 96(%r9), %xmm2
+; AVX512-FCP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm10 = [0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
+; AVX512-FCP-NEXT: vpshufb %xmm10, %xmm2, %xmm0
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,2,1,4,5,6,7]
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm26 = [0,0,0,0,8,8,0,9]
+; AVX512-FCP-NEXT: vpermi2q %zmm0, %zmm2, %zmm26
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm0 = ymm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm3[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm4[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11]
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm4[4],ymm3[4],ymm4[5],ymm3[5],ymm4[6],ymm3[6],ymm4[7],ymm3[7],ymm4[12],ymm3[12],ymm4[13],ymm3[13],ymm4[14],ymm3[14],ymm4[15],ymm3[15]
+; AVX512-FCP-NEXT: vpermt2d %zmm2, %zmm22, %zmm0
; AVX512-FCP-NEXT: vmovdqa 64(%rsi), %ymm4
; AVX512-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm4, %ymm1
+; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm4, %ymm2
; AVX512-FCP-NEXT: vmovdqa 64(%rdi), %ymm5
; AVX512-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm5, %ymm3
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm3[0],ymm1[0],ymm3[1],ymm1[1],ymm3[2],ymm1[2],ymm3[3],ymm1[3],ymm3[8],ymm1[8],ymm3[9],ymm1[9],ymm3[10],ymm1[10],ymm3[11],ymm1[11]
+; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm5, %ymm3
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm5[4],ymm4[4],ymm5[5],ymm4[5],ymm5[6],ymm4[6],ymm5[7],ymm4[7],ymm5[12],ymm4[12],ymm5[13],ymm4[13],ymm5[14],ymm4[14],ymm5[15],ymm4[15]
-; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm23, %zmm1
-; AVX512-FCP-NEXT: vmovdqa32 %zmm2, %zmm1 {%k1}
-; AVX512-FCP-NEXT: vmovdqa %ymm1, %ymm2
+; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm23, %zmm2
+; AVX512-FCP-NEXT: vmovdqa32 %zmm0, %zmm2 {%k1}
+; AVX512-FCP-NEXT: vmovdqa %ymm2, %ymm0
; AVX512-FCP-NEXT: vmovdqa 64(%r8), %ymm4
; AVX512-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm3 = ymm4[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
-; AVX512-FCP-NEXT: vpermt2d %ymm3, %ymm21, %ymm2
-; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm4, %ymm3
-; AVX512-FCP-NEXT: vpermt2d %zmm3, %zmm24, %zmm1
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 64(%r9), %ymm2
-; AVX512-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm2, %ymm1
-; AVX512-FCP-NEXT: vpshufb %ymm15, %ymm2, %ymm2
-; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm25, %zmm2
-; AVX512-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 96(%rcx), %xmm3
-; AVX512-FCP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 96(%rdx), %xmm4
+; AVX512-FCP-NEXT: vpermt2d %ymm3, %ymm21, %ymm0
+; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm4, %ymm3
+; AVX512-FCP-NEXT: vpermt2d %zmm3, %zmm24, %zmm2
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm0 = [2,2,2,3,10,9,10,11]
+; AVX512-FCP-NEXT: vmovdqa 64(%r9), %ymm3
+; AVX512-FCP-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm3, %ymm2
+; AVX512-FCP-NEXT: vpshufb %ymm15, %ymm3, %ymm3
+; AVX512-FCP-NEXT: vpermt2q %zmm2, %zmm0, %zmm3
+; AVX512-FCP-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-FCP-NEXT: vmovdqa 96(%rcx), %xmm4
; AVX512-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vmovdqa 96(%rdx), %xmm5
+; AVX512-FCP-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm2 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm26 = [1,0,2,2,1,0,2,2,16,17,16,17,16,17,16,17]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm26, %zmm2
-; AVX512-FCP-NEXT: vmovdqa 96(%rsi), %xmm4
-; AVX512-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm8 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
-; AVX512-FCP-NEXT: vpshufb %xmm8, %xmm4, %xmm1
-; AVX512-FCP-NEXT: vmovdqa 96(%rdi), %xmm5
-; AVX512-FCP-NEXT: vmovdqa %xmm5, (%rsp) # 16-byte Spill
-; AVX512-FCP-NEXT: vpshufb %xmm8, %xmm5, %xmm3
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm27 = [8,8,0,9,0,1,0,1]
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm3 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm30 = [1,0,2,2,1,0,2,2,16,17,16,17,16,17,16,17]
; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm27, %zmm1
+; AVX512-FCP-NEXT: vpermt2d %zmm2, %zmm30, %zmm3
+; AVX512-FCP-NEXT: vmovdqa 96(%rsi), %xmm6
+; AVX512-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
+; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm6, %xmm4
+; AVX512-FCP-NEXT: vmovdqa 96(%rdi), %xmm7
+; AVX512-FCP-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm7, %xmm5
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm29 = [8,8,0,9,0,1,0,1]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm29, %zmm4
; AVX512-FCP-NEXT: movw $9362, %ax # imm = 0x2492
; AVX512-FCP-NEXT: kmovw %eax, %k2
-; AVX512-FCP-NEXT: vmovdqa32 %zmm2, %zmm1 {%k2}
-; AVX512-FCP-NEXT: vmovdqa %ymm1, %ymm2
+; AVX512-FCP-NEXT: vmovdqa32 %zmm3, %zmm4 {%k2}
+; AVX512-FCP-NEXT: vmovdqa %ymm4, %ymm3
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm28 = [16,9,10,17,12,13,18,15]
-; AVX512-FCP-NEXT: vmovdqa 96(%r8), %xmm4
-; AVX512-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm4[2,1,3,3,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2d %zmm3, %zmm28, %zmm1
-; AVX512-FCP-NEXT: vpmovzxwd {{.*#+}} xmm3 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm29 = [0,1,8,3,4,9,6,7]
-; AVX512-FCP-NEXT: vpermt2d %ymm3, %ymm29, %ymm2
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm30 = [0,0,0,0,8,8,0,9]
-; AVX512-FCP-NEXT: vmovdqa 96(%r9), %xmm2
-; AVX512-FCP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm9 = [0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
-; AVX512-FCP-NEXT: vpshufb %xmm9, %xmm2, %xmm1
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,2,1,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm30, %zmm2
-; AVX512-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 64(%rcx), %xmm3
-; AVX512-FCP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 64(%rdx), %xmm4
-; AVX512-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm2 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm26, %zmm2
-; AVX512-FCP-NEXT: vmovdqa 64(%rsi), %xmm4
-; AVX512-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpshufb %xmm8, %xmm4, %xmm1
-; AVX512-FCP-NEXT: vmovdqa 64(%rdi), %xmm5
+; AVX512-FCP-NEXT: vmovdqa 96(%r8), %xmm6
+; AVX512-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm6[2,1,3,3,4,5,6,7]
+; AVX512-FCP-NEXT: vpermt2d %zmm5, %zmm28, %zmm4
+; AVX512-FCP-NEXT: vpmovzxwd {{.*#+}} xmm5 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm27 = [0,1,8,3,4,9,6,7]
+; AVX512-FCP-NEXT: vpermt2d %ymm5, %ymm27, %ymm3
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm3, %zmm3
+; AVX512-FCP-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-FCP-NEXT: vmovdqa 64(%rcx), %xmm5
; AVX512-FCP-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpshufb %xmm8, %xmm5, %xmm3
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm27, %zmm1
-; AVX512-FCP-NEXT: vmovdqa32 %zmm2, %zmm1 {%k2}
-; AVX512-FCP-NEXT: vmovdqa %ymm1, %ymm2
-; AVX512-FCP-NEXT: vmovdqa 64(%r8), %xmm4
+; AVX512-FCP-NEXT: vmovdqa 64(%rdx), %xmm6
+; AVX512-FCP-NEXT: vmovdqa %xmm6, (%rsp) # 16-byte Spill
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm3 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm4 = xmm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX512-FCP-NEXT: vpermt2d %zmm3, %zmm30, %zmm4
+; AVX512-FCP-NEXT: vmovdqa 64(%rsi), %xmm6
+; AVX512-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm6, %xmm3
+; AVX512-FCP-NEXT: vmovdqa 64(%rdi), %xmm7
+; AVX512-FCP-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm7, %xmm5
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm29, %zmm3
+; AVX512-FCP-NEXT: vmovdqa32 %zmm4, %zmm3 {%k2}
+; AVX512-FCP-NEXT: vmovdqa %ymm3, %ymm4
+; AVX512-FCP-NEXT: vmovdqa 64(%r8), %xmm6
+; AVX512-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm6[2,1,3,3,4,5,6,7]
+; AVX512-FCP-NEXT: vpermt2d %zmm5, %zmm28, %zmm3
+; AVX512-FCP-NEXT: vpmovzxwd {{.*#+}} xmm5 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
+; AVX512-FCP-NEXT: vpermt2d %ymm5, %ymm27, %ymm4
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm3, %zmm4, %zmm3
+; AVX512-FCP-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm31 = [0,0,0,0,8,8,0,9]
+; AVX512-FCP-NEXT: vmovdqa 64(%r9), %xmm4
; AVX512-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm4[2,1,3,3,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2d %zmm3, %zmm28, %zmm1
-; AVX512-FCP-NEXT: vpmovzxwd {{.*#+}} xmm3 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
-; AVX512-FCP-NEXT: vpermt2d %ymm3, %ymm29, %ymm2
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 64(%r9), %xmm2
-; AVX512-FCP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpshufb %xmm9, %xmm2, %xmm1
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,2,1,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm30, %zmm2
-; AVX512-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 32(%rcx), %ymm3
-; AVX512-FCP-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %ymm4
-; AVX512-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm1 = ymm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm3[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm4[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm4[4],ymm3[4],ymm4[5],ymm3[5],ymm4[6],ymm3[6],ymm4[7],ymm3[7],ymm4[12],ymm3[12],ymm4[13],ymm3[13],ymm4[14],ymm3[14],ymm4[15],ymm3[15]
-; AVX512-FCP-NEXT: vpermt2d %zmm2, %zmm22, %zmm1
-; AVX512-FCP-NEXT: vmovdqa 32(%rsi), %ymm4
-; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm4, %ymm2
-; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %ymm5
+; AVX512-FCP-NEXT: vpshufb %xmm10, %xmm4, %xmm3
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,0,2,1,4,5,6,7]
+; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm31, %zmm4
+; AVX512-FCP-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-FCP-NEXT: vmovdqa 32(%rcx), %ymm5
; AVX512-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm5, %ymm3
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm5[4],ymm4[4],ymm5[5],ymm4[5],ymm5[6],ymm4[6],ymm5[7],ymm4[7],ymm5[12],ymm4[12],ymm5[13],ymm4[13],ymm5[14],ymm4[14],ymm5[15],ymm4[15]
-; AVX512-FCP-NEXT: vmovdqa64 %ymm4, %ymm31
-; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm23, %zmm2
-; AVX512-FCP-NEXT: vmovdqa32 %zmm1, %zmm2 {%k1}
-; AVX512-FCP-NEXT: vmovdqa %ymm2, %ymm1
-; AVX512-FCP-NEXT: vmovdqa 32(%r8), %ymm4
+; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %ymm6
+; AVX512-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm3 = ymm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm5[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm4 = ymm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm6[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[8],ymm3[8],ymm4[9],ymm3[9],ymm4[10],ymm3[10],ymm4[11],ymm3[11]
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm6[4],ymm5[4],ymm6[5],ymm5[5],ymm6[6],ymm5[6],ymm6[7],ymm5[7],ymm6[12],ymm5[12],ymm6[13],ymm5[13],ymm6[14],ymm5[14],ymm6[15],ymm5[15]
+; AVX512-FCP-NEXT: vpermt2d %zmm4, %zmm22, %zmm3
+; AVX512-FCP-NEXT: vmovdqa 32(%rsi), %ymm6
+; AVX512-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm6, %ymm4
+; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %ymm7
+; AVX512-FCP-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm7, %ymm5
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm4 = ymm5[0],ymm4[0],ymm5[1],ymm4[1],ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[8],ymm4[8],ymm5[9],ymm4[9],ymm5[10],ymm4[10],ymm5[11],ymm4[11]
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm5 = ymm7[4],ymm6[4],ymm7[5],ymm6[5],ymm7[6],ymm6[6],ymm7[7],ymm6[7],ymm7[12],ymm6[12],ymm7[13],ymm6[13],ymm7[14],ymm6[14],ymm7[15],ymm6[15]
+; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm23, %zmm4
+; AVX512-FCP-NEXT: vmovdqa32 %zmm3, %zmm4 {%k1}
+; AVX512-FCP-NEXT: vmovdqa %ymm4, %ymm3
+; AVX512-FCP-NEXT: vmovdqa 32(%r8), %ymm6
+; AVX512-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm5 = ymm6[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
+; AVX512-FCP-NEXT: vpermt2d %ymm5, %ymm21, %ymm3
+; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm6, %ymm5
+; AVX512-FCP-NEXT: vpermt2d %zmm5, %zmm24, %zmm4
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm3, %zmm3
+; AVX512-FCP-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-FCP-NEXT: vmovdqa 32(%r9), %ymm4
; AVX512-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm3 = ymm4[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
-; AVX512-FCP-NEXT: vpermt2d %ymm3, %ymm21, %ymm1
-; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm4, %ymm3
-; AVX512-FCP-NEXT: vpermt2d %zmm3, %zmm24, %zmm2
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 32(%r9), %ymm2
-; AVX512-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm2, %ymm1
-; AVX512-FCP-NEXT: vpshufb %ymm15, %ymm2, %ymm2
-; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm25, %zmm2
-; AVX512-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm4, %ymm3
+; AVX512-FCP-NEXT: vpshufb %ymm15, %ymm4, %ymm4
+; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm0, %zmm4
+; AVX512-FCP-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT: vmovdqa (%rcx), %ymm14
; AVX512-FCP-NEXT: vmovdqa (%rdx), %ymm13
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm1 = ymm14[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm14[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm13[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm13[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm13[4],ymm14[4],ymm13[5],ymm14[5],ymm13[6],ymm14[6],ymm13[7],ymm14[7],ymm13[12],ymm14[12],ymm13[13],ymm14[13],ymm13[14],ymm14[14],ymm13[15],ymm14[15]
-; AVX512-FCP-NEXT: vpermt2d %zmm2, %zmm22, %zmm1
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm3 = ymm14[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm14[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} ymm4 = ymm13[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm13[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[8],ymm3[8],ymm4[9],ymm3[9],ymm4[10],ymm3[10],ymm4[11],ymm3[11]
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm13[4],ymm14[4],ymm13[5],ymm14[5],ymm13[6],ymm14[6],ymm13[7],ymm14[7],ymm13[12],ymm14[12],ymm13[13],ymm14[13],ymm13[14],ymm14[14],ymm13[15],ymm14[15]
+; AVX512-FCP-NEXT: vpermt2d %zmm4, %zmm22, %zmm3
; AVX512-FCP-NEXT: vmovdqa (%rsi), %ymm10
-; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm10, %ymm2
-; AVX512-FCP-NEXT: vmovdqa (%rdi), %ymm9
-; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm9, %ymm0
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[1],ymm2[1],ymm0[2],ymm2[2],ymm0[3],ymm2[3],ymm0[8],ymm2[8],ymm0[9],ymm2[9],ymm0[10],ymm2[10],ymm0[11],ymm2[11]
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm9[4],ymm10[4],ymm9[5],ymm10[5],ymm9[6],ymm10[6],ymm9[7],ymm10[7],ymm9[12],ymm10[12],ymm9[13],ymm10[13],ymm9[14],ymm10[14],ymm9[15],ymm10[15]
-; AVX512-FCP-NEXT: vpermt2q %zmm2, %zmm23, %zmm0
-; AVX512-FCP-NEXT: vmovdqa32 %zmm1, %zmm0 {%k1}
+; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm10, %ymm4
+; AVX512-FCP-NEXT: vmovdqa (%rdi), %ymm7
+; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm7, %ymm1
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[1],ymm4[1],ymm1[2],ymm4[2],ymm1[3],ymm4[3],ymm1[8],ymm4[8],ymm1[9],ymm4[9],ymm1[10],ymm4[10],ymm1[11],ymm4[11]
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm7[4],ymm10[4],ymm7[5],ymm10[5],ymm7[6],ymm10[6],ymm7[7],ymm10[7],ymm7[12],ymm10[12],ymm7[13],ymm10[13],ymm7[14],ymm10[14],ymm7[15],ymm10[15]
+; AVX512-FCP-NEXT: vpermt2q %zmm4, %zmm23, %zmm1
+; AVX512-FCP-NEXT: vmovdqa32 %zmm3, %zmm1 {%k1}
; AVX512-FCP-NEXT: vmovdqa (%r8), %ymm12
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm1 = ymm12[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
-; AVX512-FCP-NEXT: vpermi2d %ymm1, %ymm0, %ymm21
-; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm12, %ymm1
-; AVX512-FCP-NEXT: vpermt2d %zmm1, %zmm24, %zmm0
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm21, %zmm21
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm3 = ymm12[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
+; AVX512-FCP-NEXT: vpermi2d %ymm3, %ymm1, %ymm21
+; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm12, %ymm3
+; AVX512-FCP-NEXT: vpermt2d %zmm3, %zmm24, %zmm1
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm21, %zmm25
; AVX512-FCP-NEXT: vmovdqa (%r9), %ymm11
-; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm11, %ymm0
-; AVX512-FCP-NEXT: vpshufb %ymm15, %ymm11, %ymm1
-; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm25, %zmm1
-; AVX512-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 32(%rcx), %xmm2
-; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %xmm3
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm0 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm11, %ymm1
+; AVX512-FCP-NEXT: vpshufb %ymm15, %ymm11, %ymm3
+; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm0, %zmm3
+; AVX512-FCP-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-FCP-NEXT: vmovdqa 32(%rcx), %xmm3
+; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %xmm4
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm0 = xmm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm3, %xmm20
-; AVX512-FCP-NEXT: vmovdqa64 %xmm2, %xmm19
-; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm26, %zmm1
-; AVX512-FCP-NEXT: vmovdqa 32(%rsi), %xmm3
-; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %xmm4
-; AVX512-FCP-NEXT: vpshufb %xmm8, %xmm3, %xmm0
-; AVX512-FCP-NEXT: vpshufb %xmm8, %xmm4, %xmm2
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm4, %xmm18
-; AVX512-FCP-NEXT: vmovdqa64 %xmm3, %xmm17
-; AVX512-FCP-NEXT: vpermt2q %zmm2, %zmm27, %zmm0
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX512-FCP-NEXT: vmovdqa64 %xmm4, %xmm21
+; AVX512-FCP-NEXT: vmovdqa64 %xmm3, %xmm19
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm30, %zmm1
+; AVX512-FCP-NEXT: vmovdqa 32(%rsi), %xmm4
+; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %xmm5
+; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm4, %xmm0
+; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm5, %xmm3
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; AVX512-FCP-NEXT: vmovdqa64 %xmm5, %xmm18
+; AVX512-FCP-NEXT: vmovdqa64 %xmm4, %xmm17
+; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm29, %zmm0
; AVX512-FCP-NEXT: vmovdqa32 %zmm1, %zmm0 {%k2}
; AVX512-FCP-NEXT: vmovdqa %ymm0, %ymm1
-; AVX512-FCP-NEXT: vmovdqa 32(%r8), %xmm3
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm3[2,1,3,3,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2d %zmm2, %zmm28, %zmm0
-; AVX512-FCP-NEXT: vpmovzxwd {{.*#+}} xmm2 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
-; AVX512-FCP-NEXT: vmovdqa64 %xmm3, %xmm16
-; AVX512-FCP-NEXT: vpermt2d %ymm2, %ymm29, %ymm1
+; AVX512-FCP-NEXT: vmovdqa 32(%r8), %xmm4
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm4[2,1,3,3,4,5,6,7]
+; AVX512-FCP-NEXT: vpermt2d %zmm3, %zmm28, %zmm0
+; AVX512-FCP-NEXT: vpmovzxwd {{.*#+}} xmm3 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
+; AVX512-FCP-NEXT: vmovdqa64 %xmm4, %xmm16
+; AVX512-FCP-NEXT: vpermt2d %ymm3, %ymm27, %ymm1
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm22
; AVX512-FCP-NEXT: vmovdqa (%rcx), %xmm6
; AVX512-FCP-NEXT: vmovdqa (%rdx), %xmm5
; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm0 = xmm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX512-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
-; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm26, %zmm2
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm30, %zmm3
; AVX512-FCP-NEXT: vmovdqa (%rsi), %xmm4
; AVX512-FCP-NEXT: vmovdqa (%rdi), %xmm1
-; AVX512-FCP-NEXT: vpshufb %xmm8, %xmm4, %xmm3
-; AVX512-FCP-NEXT: vpshufb %xmm8, %xmm1, %xmm7
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm7[4],xmm3[4],xmm7[5],xmm3[5],xmm7[6],xmm3[6],xmm7[7],xmm3[7]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
-; AVX512-FCP-NEXT: vpermt2q %zmm7, %zmm27, %zmm3
-; AVX512-FCP-NEXT: vmovdqa32 %zmm2, %zmm3 {%k2}
+; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm4, %xmm8
+; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm1, %xmm2
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm8 = xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; AVX512-FCP-NEXT: vpermt2q %zmm2, %zmm29, %zmm8
+; AVX512-FCP-NEXT: vmovdqa32 %zmm3, %zmm8 {%k2}
; AVX512-FCP-NEXT: vmovdqa (%r8), %xmm2
-; AVX512-FCP-NEXT: vpmovzxwd {{.*#+}} xmm7 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
-; AVX512-FCP-NEXT: vpermi2d %ymm7, %ymm3, %ymm29
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm7 = xmm2[2,1,3,3,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2d %zmm7, %zmm28, %zmm3
+; AVX512-FCP-NEXT: vpmovzxwd {{.*#+}} xmm3 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
+; AVX512-FCP-NEXT: vpermi2d %ymm3, %ymm8, %ymm27
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,1,3,3,4,5,6,7]
+; AVX512-FCP-NEXT: vpermt2d %zmm3, %zmm28, %zmm8
; AVX512-FCP-NEXT: vmovdqa 32(%r9), %xmm15
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
-; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm15, %xmm7
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm8 = xmm15[0,0,2,1,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2q %zmm7, %zmm30, %zmm8
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm3, %zmm29, %zmm23
+; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm15, %xmm3
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm9 = xmm15[0,0,2,1,4,5,6,7]
+; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm31, %zmm9
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm27, %zmm23
; AVX512-FCP-NEXT: vmovdqa (%r9), %xmm3
; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm3, %xmm0
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm7 = xmm3[0,0,2,1,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm30, %zmm7
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm9[0],ymm10[0],ymm9[1],ymm10[1],ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[8],ymm10[8],ymm9[9],ymm10[9],ymm9[10],ymm10[10],ymm9[11],ymm10[11]
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm8 = xmm3[0,0,2,1,4,5,6,7]
+; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm31, %zmm8
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm7[0],ymm10[0],ymm7[1],ymm10[1],ymm7[2],ymm10[2],ymm7[3],ymm10[3],ymm7[8],ymm10[8],ymm7[9],ymm10[9],ymm7[10],ymm10[10],ymm7[11],ymm10[11]
; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm24 = [1,1,1,1,10,10,10,11]
; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm24, %zmm1
; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm4 = ymm13[0],ymm14[0],ymm13[1],ymm14[1],ymm13[2],ymm14[2],ymm13[3],ymm14[3],ymm13[8],ymm14[8],ymm13[9],ymm14[9],ymm13[10],ymm14[10],ymm13[11],ymm14[11]
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm28 = [17,18,17,18,0,0,19,19,5,4,2,2,5,4,6,6]
-; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm28, %zmm4
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm30 = [17,18,17,18,0,0,19,19,5,4,2,2,5,4,6,6]
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm30, %zmm4
; AVX512-FCP-NEXT: vmovdqa32 %zmm1, %zmm4 {%k1}
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm9 = [12,13,10,11,10,11,14,15,14,15,14,15,14,15,14,15]
-; AVX512-FCP-NEXT: vpshufb %xmm9, %xmm2, %xmm0
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm5 = [0,9,2,3,8,5,6,11]
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm5 = [12,13,10,11,10,11,14,15,14,15,14,15,14,15,14,15]
+; AVX512-FCP-NEXT: vpshufb %xmm5, %xmm2, %xmm0
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm6 = [0,9,2,3,8,5,6,11]
; AVX512-FCP-NEXT: vmovdqa %ymm4, %ymm1
-; AVX512-FCP-NEXT: vpermt2d %ymm0, %ymm5, %ymm1
+; AVX512-FCP-NEXT: vpermt2d %ymm0, %ymm6, %ymm1
; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm0 = ymm12[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm25 = [8,9,20,11,12,21,14,15]
-; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm25, %zmm4
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm1, %zmm26
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm27 = [8,9,20,11,12,21,14,15]
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm27, %zmm4
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm1, %zmm28
; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm0 = ymm11[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm4 = [8,9,12,13,12,13,10,11,8,9,10,11,12,13,14,15]
; AVX512-FCP-NEXT: vpshufb %xmm4, %xmm3, %xmm2
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm1 = [0,1,0,1,10,10,10,10]
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm1 = [0,0,0,1,0,10,10,0]
; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm1, %zmm2
; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX512-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX512-FCP-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11]
-; AVX512-FCP-NEXT: vmovdqa (%rsp), %xmm3 # 16-byte Reload
+; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX512-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX512-FCP-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm24, %zmm3
; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512-FCP-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX512-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX512-FCP-NEXT: # ymm6 = ymm6[0],mem[0],ymm6[1],mem[1],ymm6[2],mem[2],ymm6[3],mem[3],ymm6[8],mem[8],ymm6[9],mem[9],ymm6[10],mem[10],ymm6[11],mem[11]
-; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm28, %zmm6
-; AVX512-FCP-NEXT: vmovdqa32 %zmm3, %zmm6 {%k1}
+; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX512-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: # ymm7 = ymm7[0],mem[0],ymm7[1],mem[1],ymm7[2],mem[2],ymm7[3],mem[3],ymm7[8],mem[8],ymm7[9],mem[9],ymm7[10],mem[10],ymm7[11],mem[11]
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm30, %zmm7
+; AVX512-FCP-NEXT: vmovdqa32 %zmm3, %zmm7 {%k1}
; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX512-FCP-NEXT: vpshufb %xmm9, %xmm0, %xmm0
-; AVX512-FCP-NEXT: vmovdqa %ymm6, %ymm3
-; AVX512-FCP-NEXT: vpermt2d %ymm0, %ymm5, %ymm3
+; AVX512-FCP-NEXT: vpshufb %xmm5, %xmm0, %xmm0
+; AVX512-FCP-NEXT: vmovdqa %ymm7, %ymm3
+; AVX512-FCP-NEXT: vpermt2d %ymm0, %ymm6, %ymm3
; AVX512-FCP-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512-FCP-NEXT: # ymm0 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm25, %zmm6
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm6, %zmm3, %zmm27
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm27, %zmm7
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm7, %zmm3, %zmm29
; AVX512-FCP-NEXT: vpshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512-FCP-NEXT: # ymm0 = mem[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX512-FCP-NEXT: vpshufb %xmm4, %xmm3, %xmm6
-; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm1, %zmm6
+; AVX512-FCP-NEXT: vpshufb %xmm4, %xmm3, %xmm7
+; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm1, %zmm7
; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX512-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX512-FCP-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11]
@@ -9082,71 +9078,71 @@ define void @store_i16_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX512-FCP-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; AVX512-FCP-NEXT: vpermt2q %zmm0, %zmm24, %zmm3
-; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX512-FCP-NEXT: vmovdqa (%rsp), %xmm0 # 16-byte Reload
; AVX512-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512-FCP-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX512-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX512-FCP-NEXT: # ymm10 = ymm10[0],mem[0],ymm10[1],mem[1],ymm10[2],mem[2],ymm10[3],mem[3],ymm10[8],mem[8],ymm10[9],mem[9],ymm10[10],mem[10],ymm10[11],mem[11]
-; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm28, %zmm10
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm30, %zmm10
; AVX512-FCP-NEXT: vmovdqa32 %zmm3, %zmm10 {%k1}
; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX512-FCP-NEXT: vpshufb %xmm9, %xmm0, %xmm0
+; AVX512-FCP-NEXT: vpshufb %xmm5, %xmm0, %xmm0
; AVX512-FCP-NEXT: vmovdqa %ymm10, %ymm3
-; AVX512-FCP-NEXT: vpermt2d %ymm0, %ymm5, %ymm3
+; AVX512-FCP-NEXT: vpermt2d %ymm0, %ymm6, %ymm3
; AVX512-FCP-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512-FCP-NEXT: # ymm0 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm25, %zmm10
+; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm27, %zmm10
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm10, %zmm3, %zmm3
; AVX512-FCP-NEXT: vpshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
; AVX512-FCP-NEXT: # ymm10 = mem[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512-FCP-NEXT: vpshufb %xmm4, %xmm0, %xmm0
; AVX512-FCP-NEXT: vpermt2q %zmm10, %zmm1, %zmm0
-; AVX512-FCP-NEXT: vmovdqa64 %ymm31, %ymm11
; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm10 = ymm10[0],ymm11[0],ymm10[1],ymm11[1],ymm10[2],ymm11[2],ymm10[3],ymm11[3],ymm10[8],ymm11[8],ymm10[9],ymm11[9],ymm10[10],ymm11[10],ymm10[11],ymm11[11]
+; AVX512-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: # ymm10 = ymm10[0],mem[0],ymm10[1],mem[1],ymm10[2],mem[2],ymm10[3],mem[3],ymm10[8],mem[8],ymm10[9],mem[9],ymm10[10],mem[10],ymm10[11],mem[11]
; AVX512-FCP-NEXT: vmovdqa64 %xmm18, %xmm11
; AVX512-FCP-NEXT: vmovdqa64 %xmm17, %xmm12
; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm11 = xmm11[4],xmm12[4],xmm11[5],xmm12[5],xmm11[6],xmm12[6],xmm11[7],xmm12[7]
; AVX512-FCP-NEXT: vpermt2q %zmm10, %zmm24, %zmm11
-; AVX512-FCP-NEXT: vmovdqa64 %xmm20, %xmm10
+; AVX512-FCP-NEXT: vmovdqa64 %xmm21, %xmm10
; AVX512-FCP-NEXT: vmovdqa64 %xmm19, %xmm12
; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm10 = xmm10[4],xmm12[4],xmm10[5],xmm12[5],xmm10[6],xmm12[6],xmm10[7],xmm12[7]
; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX512-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
; AVX512-FCP-NEXT: # ymm12 = ymm12[0],mem[0],ymm12[1],mem[1],ymm12[2],mem[2],ymm12[3],mem[3],ymm12[8],mem[8],ymm12[9],mem[9],ymm12[10],mem[10],ymm12[11],mem[11]
-; AVX512-FCP-NEXT: vpermt2d %zmm10, %zmm28, %zmm12
+; AVX512-FCP-NEXT: vpermt2d %zmm10, %zmm30, %zmm12
; AVX512-FCP-NEXT: vmovdqa32 %zmm11, %zmm12 {%k1}
; AVX512-FCP-NEXT: vmovdqa64 %xmm16, %xmm10
-; AVX512-FCP-NEXT: vpshufb %xmm9, %xmm10, %xmm9
-; AVX512-FCP-NEXT: vpermi2d %ymm9, %ymm12, %ymm5
-; AVX512-FCP-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
-; AVX512-FCP-NEXT: # ymm9 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX512-FCP-NEXT: vpermt2d %zmm9, %zmm25, %zmm12
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm12, %zmm5, %zmm5
+; AVX512-FCP-NEXT: vpshufb %xmm5, %xmm10, %xmm5
+; AVX512-FCP-NEXT: vpermi2d %ymm5, %ymm12, %ymm6
+; AVX512-FCP-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: # ymm5 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
+; AVX512-FCP-NEXT: vpermt2d %zmm5, %zmm27, %zmm12
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm12, %zmm6, %zmm5
; AVX512-FCP-NEXT: vpshufb %xmm4, %xmm15, %xmm4
-; AVX512-FCP-NEXT: vpshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
-; AVX512-FCP-NEXT: # ymm9 = mem[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
-; AVX512-FCP-NEXT: vpermt2q %zmm9, %zmm1, %zmm4
+; AVX512-FCP-NEXT: vpshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: # ymm6 = mem[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
+; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm1, %zmm4
; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm4 ^ (zmm1 & (zmm4 ^ zmm5))
; AVX512-FCP-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512-FCP-NEXT: vmovdqa64 %zmm4, 256(%rax)
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (zmm1 & (zmm0 ^ zmm3))
; AVX512-FCP-NEXT: vmovdqa64 %zmm0, 448(%rax)
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm6 ^ (zmm1 & (zmm6 ^ zmm27))
-; AVX512-FCP-NEXT: vmovdqa64 %zmm6, 640(%rax)
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm1 & (zmm2 ^ zmm26))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (zmm1 & (zmm7 ^ zmm29))
+; AVX512-FCP-NEXT: vmovdqa64 %zmm7, 640(%rax)
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm1 & (zmm2 ^ zmm28))
; AVX512-FCP-NEXT: vmovdqa64 %zmm2, 64(%rax)
; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (zmm0 & (zmm7 ^ zmm23))
-; AVX512-FCP-NEXT: vmovdqa64 %zmm7, (%rax)
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm8 ^ (zmm0 & (zmm8 ^ zmm22))
-; AVX512-FCP-NEXT: vmovdqa64 %zmm8, 192(%rax)
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm8 ^ (zmm0 & (zmm8 ^ zmm23))
+; AVX512-FCP-NEXT: vmovdqa64 %zmm8, (%rax)
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm9 = zmm9 ^ (zmm0 & (zmm9 ^ zmm22))
+; AVX512-FCP-NEXT: vmovdqa64 %zmm9, 192(%rax)
; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0]
; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm1 & (zmm2 ^ zmm21))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm1 & (zmm2 ^ zmm25))
; AVX512-FCP-NEXT: vmovdqa64 %zmm2, 128(%rax)
; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm2 # 64-byte Folded Reload
@@ -9157,18 +9153,16 @@ define void @store_i16_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: # zmm2 = zmm2 ^ (zmm0 & (zmm2 ^ mem))
; AVX512-FCP-NEXT: vmovdqa64 %zmm2, 384(%rax)
; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
-; AVX512-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 # 64-byte Folded Reload
-; AVX512-FCP-NEXT: # zmm2 = zmm2 ^ (zmm0 & (zmm2 ^ mem))
-; AVX512-FCP-NEXT: vmovdqa64 %zmm2, 576(%rax)
-; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
-; AVX512-FCP-NEXT: # zmm0 = zmm0 ^ (zmm1 & (zmm0 ^ mem))
-; AVX512-FCP-NEXT: vmovdqa64 %zmm0, 512(%rax)
-; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
-; AVX512-FCP-NEXT: # zmm0 = zmm0 ^ (zmm1 & (zmm0 ^ mem))
-; AVX512-FCP-NEXT: vmovdqa64 %zmm0, 704(%rax)
-; AVX512-FCP-NEXT: addq $1320, %rsp # imm = 0x528
+; AVX512-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm2 # 64-byte Folded Reload
+; AVX512-FCP-NEXT: # zmm2 = zmm2 ^ (zmm1 & (zmm2 ^ mem))
+; AVX512-FCP-NEXT: vmovdqa64 %zmm2, 512(%rax)
+; AVX512-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm20 # 64-byte Folded Reload
+; AVX512-FCP-NEXT: # zmm20 = zmm20 ^ (zmm1 & (zmm20 ^ mem))
+; AVX512-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm26 # 64-byte Folded Reload
+; AVX512-FCP-NEXT: # zmm26 = zmm26 ^ (zmm0 & (zmm26 ^ mem))
+; AVX512-FCP-NEXT: vmovdqa64 %zmm26, 576(%rax)
+; AVX512-FCP-NEXT: vmovdqa64 %zmm20, 704(%rax)
+; AVX512-FCP-NEXT: addq $1224, %rsp # imm = 0x4C8
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
@@ -9658,7 +9652,7 @@ define void @store_i16_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX512DQ-FCP-LABEL: store_i16_stride6_vf64:
; AVX512DQ-FCP: # %bb.0:
-; AVX512DQ-FCP-NEXT: subq $1320, %rsp # imm = 0x528
+; AVX512DQ-FCP-NEXT: subq $1224, %rsp # imm = 0x4C8
; AVX512DQ-FCP-NEXT: vmovdqa 96(%rcx), %ymm1
; AVX512DQ-FCP-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa 96(%rdx), %ymm2
@@ -9666,324 +9660,324 @@ define void @store_i16_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm2[4],ymm1[4],ymm2[5],ymm1[5],ymm2[6],ymm1[6],ymm2[7],ymm1[7],ymm2[12],ymm1[12],ymm2[13],ymm1[13],ymm2[14],ymm1[14],ymm2[15],ymm1[15]
; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm1 = ymm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm1[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm2[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm22 = [4,5,4,5,4,5,4,5,21,22,21,22,21,22,23,23]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm22, %zmm1
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm22, %zmm2
; AVX512DQ-FCP-NEXT: vmovdqa 96(%rsi), %ymm4
; AVX512DQ-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm0 = [u,u,u,u,4,5,10,11,u,u,u,u,u,u,u,u,24,25,22,23,20,21,26,27,u,u,u,u,u,u,u,u]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm4, %ymm2
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm1 = [u,u,u,u,4,5,10,11,u,u,u,u,u,u,u,u,24,25,22,23,20,21,26,27,u,u,u,u,u,u,u,u]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm4, %ymm0
; AVX512DQ-FCP-NEXT: vmovdqa 96(%rdi), %ymm5
; AVX512DQ-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm5, %ymm3
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm5, %ymm3
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[1],ymm0[1],ymm3[2],ymm0[2],ymm3[3],ymm0[3],ymm3[8],ymm0[8],ymm3[9],ymm0[9],ymm3[10],ymm0[10],ymm3[11],ymm0[11]
; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm5[4],ymm4[4],ymm5[5],ymm4[5],ymm5[6],ymm4[6],ymm5[7],ymm4[7],ymm5[12],ymm4[12],ymm5[13],ymm4[13],ymm5[14],ymm4[14],ymm5[15],ymm4[15]
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm23 = [2,1,2,3,11,11,11,11]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm23, %zmm2
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm23, %zmm0
; AVX512DQ-FCP-NEXT: movw $18724, %ax # imm = 0x4924
; AVX512DQ-FCP-NEXT: kmovw %eax, %k1
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm1, %zmm2 {%k1}
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm2, %zmm0 {%k1}
; AVX512DQ-FCP-NEXT: vmovdqa 96(%r8), %ymm4
; AVX512DQ-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm1 = ymm4[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm2 = ymm4[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm21 = [12,1,2,13,4,5,14,7]
-; AVX512DQ-FCP-NEXT: vmovdqa %ymm2, %ymm3
-; AVX512DQ-FCP-NEXT: vpermt2d %ymm1, %ymm21, %ymm3
+; AVX512DQ-FCP-NEXT: vmovdqa %ymm0, %ymm3
+; AVX512DQ-FCP-NEXT: vpermt2d %ymm2, %ymm21, %ymm3
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm24 = [8,21,10,11,20,13,14,23]
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm6 = [u,u,u,u,u,u,u,u,14,15,14,15,14,15,14,15,28,29,26,27,26,27,30,31,30,31,30,31,30,31,30,31]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm4, %ymm1
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm24, %zmm2
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm8 = [u,u,u,u,u,u,u,u,14,15,14,15,14,15,14,15,28,29,26,27,26,27,30,31,30,31,30,31,30,31,30,31]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm4, %ymm2
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm2, %zmm24, %zmm0
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm3, %zmm0
+; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa 96(%r9), %ymm2
; AVX512DQ-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm7 = [u,u,u,u,u,u,u,u,8,9,10,11,12,13,14,15,24,25,28,29,28,29,26,27,24,25,26,27,28,29,30,31]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm2, %ymm1
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm9 = [u,u,u,u,u,u,u,u,8,9,10,11,12,13,14,15,24,25,28,29,28,29,26,27,24,25,26,27,28,29,30,31]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm2, %ymm0
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm15 = [16,17,20,21,20,21,22,23,24,25,24,25,24,25,24,25,16,17,20,21,20,21,22,23,24,25,24,25,24,25,24,25]
; AVX512DQ-FCP-NEXT: # ymm15 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpshufb %ymm15, %ymm2, %ymm2
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm25 = [2,2,2,3,10,9,10,11]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm25, %zmm2
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%rcx), %ymm2
-; AVX512DQ-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdx), %ymm3
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm20 = [2,2,0,3,10,0,10,11]
+; AVX512DQ-FCP-NEXT: vpermi2q %zmm0, %zmm2, %zmm20
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%rcx), %ymm3
; AVX512DQ-FCP-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm1 = ymm3[4],ymm2[4],ymm3[5],ymm2[5],ymm3[6],ymm2[6],ymm3[7],ymm2[7],ymm3[12],ymm2[12],ymm3[13],ymm2[13],ymm3[14],ymm2[14],ymm3[15],ymm2[15]
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm2[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm3 = ymm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm3[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm22, %zmm2
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdx), %ymm4
+; AVX512DQ-FCP-NEXT: vmovdqa 96(%r9), %xmm2
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm10 = [0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm10, %xmm2, %xmm0
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,2,1,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm26 = [0,0,0,0,8,8,0,9]
+; AVX512DQ-FCP-NEXT: vpermi2q %zmm0, %zmm2, %zmm26
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm0 = ymm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm3[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm4[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11]
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm4[4],ymm3[4],ymm4[5],ymm3[5],ymm4[6],ymm3[6],ymm4[7],ymm3[7],ymm4[12],ymm3[12],ymm4[13],ymm3[13],ymm4[14],ymm3[14],ymm4[15],ymm3[15]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm2, %zmm22, %zmm0
; AVX512DQ-FCP-NEXT: vmovdqa 64(%rsi), %ymm4
; AVX512DQ-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm4, %ymm1
+; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm4, %ymm2
; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdi), %ymm5
; AVX512DQ-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm5, %ymm3
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm3[0],ymm1[0],ymm3[1],ymm1[1],ymm3[2],ymm1[2],ymm3[3],ymm1[3],ymm3[8],ymm1[8],ymm3[9],ymm1[9],ymm3[10],ymm1[10],ymm3[11],ymm1[11]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm5, %ymm3
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm5[4],ymm4[4],ymm5[5],ymm4[5],ymm5[6],ymm4[6],ymm5[7],ymm4[7],ymm5[12],ymm4[12],ymm5[13],ymm4[13],ymm5[14],ymm4[14],ymm5[15],ymm4[15]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm23, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm2, %zmm1 {%k1}
-; AVX512DQ-FCP-NEXT: vmovdqa %ymm1, %ymm2
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm23, %zmm2
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm0, %zmm2 {%k1}
+; AVX512DQ-FCP-NEXT: vmovdqa %ymm2, %ymm0
; AVX512DQ-FCP-NEXT: vmovdqa 64(%r8), %ymm4
; AVX512DQ-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm3 = ymm4[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpermt2d %ymm3, %ymm21, %ymm2
-; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm4, %ymm3
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm3, %zmm24, %zmm1
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%r9), %ymm2
-; AVX512DQ-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm2, %ymm1
-; AVX512DQ-FCP-NEXT: vpshufb %ymm15, %ymm2, %ymm2
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm25, %zmm2
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 96(%rcx), %xmm3
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 96(%rdx), %xmm4
+; AVX512DQ-FCP-NEXT: vpermt2d %ymm3, %ymm21, %ymm0
+; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm4, %ymm3
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm3, %zmm24, %zmm2
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm0 = [2,2,2,3,10,9,10,11]
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%r9), %ymm3
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm3, %ymm2
+; AVX512DQ-FCP-NEXT: vpshufb %ymm15, %ymm3, %ymm3
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm2, %zmm0, %zmm3
+; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa 96(%rcx), %xmm4
; AVX512DQ-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vmovdqa 96(%rdx), %xmm5
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm2 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm26 = [1,0,2,2,1,0,2,2,16,17,16,17,16,17,16,17]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm26, %zmm2
-; AVX512DQ-FCP-NEXT: vmovdqa 96(%rsi), %xmm4
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm8 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm8, %xmm4, %xmm1
-; AVX512DQ-FCP-NEXT: vmovdqa 96(%rdi), %xmm5
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm5, (%rsp) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %xmm8, %xmm5, %xmm3
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm27 = [8,8,0,9,0,1,0,1]
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm3 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm30 = [1,0,2,2,1,0,2,2,16,17,16,17,16,17,16,17]
; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm27, %zmm1
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm2, %zmm30, %zmm3
+; AVX512DQ-FCP-NEXT: vmovdqa 96(%rsi), %xmm6
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm6, %xmm4
+; AVX512DQ-FCP-NEXT: vmovdqa 96(%rdi), %xmm7
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm7, %xmm5
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm29 = [8,8,0,9,0,1,0,1]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm29, %zmm4
; AVX512DQ-FCP-NEXT: movw $9362, %ax # imm = 0x2492
; AVX512DQ-FCP-NEXT: kmovw %eax, %k2
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm2, %zmm1 {%k2}
-; AVX512DQ-FCP-NEXT: vmovdqa %ymm1, %ymm2
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm3, %zmm4 {%k2}
+; AVX512DQ-FCP-NEXT: vmovdqa %ymm4, %ymm3
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm28 = [16,9,10,17,12,13,18,15]
-; AVX512DQ-FCP-NEXT: vmovdqa 96(%r8), %xmm4
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm4[2,1,3,3,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm3, %zmm28, %zmm1
-; AVX512DQ-FCP-NEXT: vpmovzxwd {{.*#+}} xmm3 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm29 = [0,1,8,3,4,9,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2d %ymm3, %ymm29, %ymm2
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm30 = [0,0,0,0,8,8,0,9]
-; AVX512DQ-FCP-NEXT: vmovdqa 96(%r9), %xmm2
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm9 = [0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm9, %xmm2, %xmm1
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,2,1,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm30, %zmm2
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%rcx), %xmm3
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdx), %xmm4
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm2 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm26, %zmm2
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%rsi), %xmm4
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %xmm8, %xmm4, %xmm1
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdi), %xmm5
+; AVX512DQ-FCP-NEXT: vmovdqa 96(%r8), %xmm6
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm6[2,1,3,3,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm5, %zmm28, %zmm4
+; AVX512DQ-FCP-NEXT: vpmovzxwd {{.*#+}} xmm5 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm27 = [0,1,8,3,4,9,6,7]
+; AVX512DQ-FCP-NEXT: vpermt2d %ymm5, %ymm27, %ymm3
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm3, %zmm3
+; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%rcx), %xmm5
; AVX512DQ-FCP-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %xmm8, %xmm5, %xmm3
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm27, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm2, %zmm1 {%k2}
-; AVX512DQ-FCP-NEXT: vmovdqa %ymm1, %ymm2
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%r8), %xmm4
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdx), %xmm6
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm6, (%rsp) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm3 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm4 = xmm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm3, %zmm30, %zmm4
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%rsi), %xmm6
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm6, %xmm3
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdi), %xmm7
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm7, %xmm5
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm29, %zmm3
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm4, %zmm3 {%k2}
+; AVX512DQ-FCP-NEXT: vmovdqa %ymm3, %ymm4
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%r8), %xmm6
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm6[2,1,3,3,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm5, %zmm28, %zmm3
+; AVX512DQ-FCP-NEXT: vpmovzxwd {{.*#+}} xmm5 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
+; AVX512DQ-FCP-NEXT: vpermt2d %ymm5, %ymm27, %ymm4
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm3, %zmm4, %zmm3
+; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm31 = [0,0,0,0,8,8,0,9]
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%r9), %xmm4
; AVX512DQ-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm4[2,1,3,3,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm3, %zmm28, %zmm1
-; AVX512DQ-FCP-NEXT: vpmovzxwd {{.*#+}} xmm3 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
-; AVX512DQ-FCP-NEXT: vpermt2d %ymm3, %ymm29, %ymm2
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%r9), %xmm2
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %xmm9, %xmm2, %xmm1
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,2,1,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm30, %zmm2
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rcx), %ymm3
-; AVX512DQ-FCP-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %ymm4
-; AVX512DQ-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm1 = ymm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm3[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm4[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm4[4],ymm3[4],ymm4[5],ymm3[5],ymm4[6],ymm3[6],ymm4[7],ymm3[7],ymm4[12],ymm3[12],ymm4[13],ymm3[13],ymm4[14],ymm3[14],ymm4[15],ymm3[15]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm2, %zmm22, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rsi), %ymm4
-; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm4, %ymm2
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %ymm5
+; AVX512DQ-FCP-NEXT: vpshufb %xmm10, %xmm4, %xmm3
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,0,2,1,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm31, %zmm4
+; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rcx), %ymm5
; AVX512DQ-FCP-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm5, %ymm3
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm5[4],ymm4[4],ymm5[5],ymm4[5],ymm5[6],ymm4[6],ymm5[7],ymm4[7],ymm5[12],ymm4[12],ymm5[13],ymm4[13],ymm5[14],ymm4[14],ymm5[15],ymm4[15]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm4, %ymm31
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm23, %zmm2
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm1, %zmm2 {%k1}
-; AVX512DQ-FCP-NEXT: vmovdqa %ymm2, %ymm1
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %ymm4
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %ymm6
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm3 = ymm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm5[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm4 = ymm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm6[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[8],ymm3[8],ymm4[9],ymm3[9],ymm4[10],ymm3[10],ymm4[11],ymm3[11]
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm6[4],ymm5[4],ymm6[5],ymm5[5],ymm6[6],ymm5[6],ymm6[7],ymm5[7],ymm6[12],ymm5[12],ymm6[13],ymm5[13],ymm6[14],ymm5[14],ymm6[15],ymm5[15]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm4, %zmm22, %zmm3
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rsi), %ymm6
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm6, %ymm4
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %ymm7
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm7, %ymm5
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm4 = ymm5[0],ymm4[0],ymm5[1],ymm4[1],ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[8],ymm4[8],ymm5[9],ymm4[9],ymm5[10],ymm4[10],ymm5[11],ymm4[11]
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm5 = ymm7[4],ymm6[4],ymm7[5],ymm6[5],ymm7[6],ymm6[6],ymm7[7],ymm6[7],ymm7[12],ymm6[12],ymm7[13],ymm6[13],ymm7[14],ymm6[14],ymm7[15],ymm6[15]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm23, %zmm4
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm3, %zmm4 {%k1}
+; AVX512DQ-FCP-NEXT: vmovdqa %ymm4, %ymm3
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %ymm6
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm5 = ymm6[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
+; AVX512DQ-FCP-NEXT: vpermt2d %ymm5, %ymm21, %ymm3
+; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm6, %ymm5
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm5, %zmm24, %zmm4
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm3, %zmm3
+; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%r9), %ymm4
; AVX512DQ-FCP-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm3 = ymm4[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpermt2d %ymm3, %ymm21, %ymm1
-; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm4, %ymm3
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm3, %zmm24, %zmm2
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%r9), %ymm2
-; AVX512DQ-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm2, %ymm1
-; AVX512DQ-FCP-NEXT: vpshufb %ymm15, %ymm2, %ymm2
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm25, %zmm2
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm4, %ymm3
+; AVX512DQ-FCP-NEXT: vpshufb %ymm15, %ymm4, %ymm4
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm0, %zmm4
+; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %ymm14
; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %ymm13
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm1 = ymm14[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm14[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm2 = ymm13[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm13[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm13[4],ymm14[4],ymm13[5],ymm14[5],ymm13[6],ymm14[6],ymm13[7],ymm14[7],ymm13[12],ymm14[12],ymm13[13],ymm14[13],ymm13[14],ymm14[14],ymm13[15],ymm14[15]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm2, %zmm22, %zmm1
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm3 = ymm14[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm14[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} ymm4 = ymm13[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm13[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[8],ymm3[8],ymm4[9],ymm3[9],ymm4[10],ymm3[10],ymm4[11],ymm3[11]
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm13[4],ymm14[4],ymm13[5],ymm14[5],ymm13[6],ymm14[6],ymm13[7],ymm14[7],ymm13[12],ymm14[12],ymm13[13],ymm14[13],ymm13[14],ymm14[14],ymm13[15],ymm14[15]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm4, %zmm22, %zmm3
; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %ymm10
-; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm10, %ymm2
-; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %ymm9
-; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm9, %ymm0
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[1],ymm2[1],ymm0[2],ymm2[2],ymm0[3],ymm2[3],ymm0[8],ymm2[8],ymm0[9],ymm2[9],ymm0[10],ymm2[10],ymm0[11],ymm2[11]
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm9[4],ymm10[4],ymm9[5],ymm10[5],ymm9[6],ymm10[6],ymm9[7],ymm10[7],ymm9[12],ymm10[12],ymm9[13],ymm10[13],ymm9[14],ymm10[14],ymm9[15],ymm10[15]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm2, %zmm23, %zmm0
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm1, %zmm0 {%k1}
+; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm10, %ymm4
+; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %ymm7
+; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm7, %ymm1
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[1],ymm4[1],ymm1[2],ymm4[2],ymm1[3],ymm4[3],ymm1[8],ymm4[8],ymm1[9],ymm4[9],ymm1[10],ymm4[10],ymm1[11],ymm4[11]
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm7[4],ymm10[4],ymm7[5],ymm10[5],ymm7[6],ymm10[6],ymm7[7],ymm10[7],ymm7[12],ymm10[12],ymm7[13],ymm10[13],ymm7[14],ymm10[14],ymm7[15],ymm10[15]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm4, %zmm23, %zmm1
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm3, %zmm1 {%k1}
; AVX512DQ-FCP-NEXT: vmovdqa (%r8), %ymm12
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm1 = ymm12[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpermi2d %ymm1, %ymm0, %ymm21
-; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm12, %ymm1
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm1, %zmm24, %zmm0
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm21, %zmm21
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm3 = ymm12[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
+; AVX512DQ-FCP-NEXT: vpermi2d %ymm3, %ymm1, %ymm21
+; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm12, %ymm3
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm3, %zmm24, %zmm1
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm21, %zmm25
; AVX512DQ-FCP-NEXT: vmovdqa (%r9), %ymm11
-; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm11, %ymm0
-; AVX512DQ-FCP-NEXT: vpshufb %ymm15, %ymm11, %ymm1
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm25, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rcx), %xmm2
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %xmm3
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm0 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm11, %ymm1
+; AVX512DQ-FCP-NEXT: vpshufb %ymm15, %ymm11, %ymm3
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm0, %zmm3
+; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rcx), %xmm3
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %xmm4
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm0 = xmm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm3, %xmm20
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm2, %xmm19
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm26, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rsi), %xmm3
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %xmm4
-; AVX512DQ-FCP-NEXT: vpshufb %xmm8, %xmm3, %xmm0
-; AVX512DQ-FCP-NEXT: vpshufb %xmm8, %xmm4, %xmm2
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm4, %xmm18
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm3, %xmm17
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm2, %zmm27, %zmm0
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm4, %xmm21
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm3, %xmm19
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm30, %zmm1
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rsi), %xmm4
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %xmm5
+; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm4, %xmm0
+; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm5, %xmm3
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm5, %xmm18
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm4, %xmm17
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm29, %zmm0
; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm1, %zmm0 {%k2}
; AVX512DQ-FCP-NEXT: vmovdqa %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %xmm3
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm3[2,1,3,3,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm2, %zmm28, %zmm0
-; AVX512DQ-FCP-NEXT: vpmovzxwd {{.*#+}} xmm2 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm3, %xmm16
-; AVX512DQ-FCP-NEXT: vpermt2d %ymm2, %ymm29, %ymm1
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %xmm4
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm4[2,1,3,3,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm3, %zmm28, %zmm0
+; AVX512DQ-FCP-NEXT: vpmovzxwd {{.*#+}} xmm3 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm4, %xmm16
+; AVX512DQ-FCP-NEXT: vpermt2d %ymm3, %ymm27, %ymm1
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm22
; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %xmm6
; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %xmm5
; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm0 = xmm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX512DQ-FCP-NEXT: vpsrldq {{.*#+}} xmm1 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm26, %zmm2
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm30, %zmm3
; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %xmm4
; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %xmm1
-; AVX512DQ-FCP-NEXT: vpshufb %xmm8, %xmm4, %xmm3
-; AVX512DQ-FCP-NEXT: vpshufb %xmm8, %xmm1, %xmm7
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm7[4],xmm3[4],xmm7[5],xmm3[5],xmm7[6],xmm3[6],xmm7[7],xmm3[7]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm7, %zmm27, %zmm3
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm2, %zmm3 {%k2}
+; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm4, %xmm8
+; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm1, %xmm2
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm8 = xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm2, %zmm29, %zmm8
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm3, %zmm8 {%k2}
; AVX512DQ-FCP-NEXT: vmovdqa (%r8), %xmm2
-; AVX512DQ-FCP-NEXT: vpmovzxwd {{.*#+}} xmm7 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
-; AVX512DQ-FCP-NEXT: vpermi2d %ymm7, %ymm3, %ymm29
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm7 = xmm2[2,1,3,3,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm7, %zmm28, %zmm3
+; AVX512DQ-FCP-NEXT: vpmovzxwd {{.*#+}} xmm3 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
+; AVX512DQ-FCP-NEXT: vpermi2d %ymm3, %ymm8, %ymm27
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,1,3,3,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm3, %zmm28, %zmm8
; AVX512DQ-FCP-NEXT: vmovdqa 32(%r9), %xmm15
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm15, %xmm7
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm8 = xmm15[0,0,2,1,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm7, %zmm30, %zmm8
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm3, %zmm29, %zmm23
+; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm15, %xmm3
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm9 = xmm15[0,0,2,1,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm31, %zmm9
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm27, %zmm23
; AVX512DQ-FCP-NEXT: vmovdqa (%r9), %xmm3
; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm3, %xmm0
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm7 = xmm3[0,0,2,1,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm30, %zmm7
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm9[0],ymm10[0],ymm9[1],ymm10[1],ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[8],ymm10[8],ymm9[9],ymm10[9],ymm9[10],ymm10[10],ymm9[11],ymm10[11]
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm8 = xmm3[0,0,2,1,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm31, %zmm8
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm7[0],ymm10[0],ymm7[1],ymm10[1],ymm7[2],ymm10[2],ymm7[3],ymm10[3],ymm7[8],ymm10[8],ymm7[9],ymm10[9],ymm7[10],ymm10[10],ymm7[11],ymm10[11]
; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm24 = [1,1,1,1,10,10,10,11]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm24, %zmm1
; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm4 = ymm13[0],ymm14[0],ymm13[1],ymm14[1],ymm13[2],ymm14[2],ymm13[3],ymm14[3],ymm13[8],ymm14[8],ymm13[9],ymm14[9],ymm13[10],ymm14[10],ymm13[11],ymm14[11]
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm28 = [17,18,17,18,0,0,19,19,5,4,2,2,5,4,6,6]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm28, %zmm4
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm30 = [17,18,17,18,0,0,19,19,5,4,2,2,5,4,6,6]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm30, %zmm4
; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm1, %zmm4 {%k1}
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm9 = [12,13,10,11,10,11,14,15,14,15,14,15,14,15,14,15]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm9, %xmm2, %xmm0
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm5 = [0,9,2,3,8,5,6,11]
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm5 = [12,13,10,11,10,11,14,15,14,15,14,15,14,15,14,15]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm5, %xmm2, %xmm0
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm6 = [0,9,2,3,8,5,6,11]
; AVX512DQ-FCP-NEXT: vmovdqa %ymm4, %ymm1
-; AVX512DQ-FCP-NEXT: vpermt2d %ymm0, %ymm5, %ymm1
+; AVX512DQ-FCP-NEXT: vpermt2d %ymm0, %ymm6, %ymm1
; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm0 = ymm12[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm25 = [8,9,20,11,12,21,14,15]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm25, %zmm4
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm1, %zmm26
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm27 = [8,9,20,11,12,21,14,15]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm27, %zmm4
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm1, %zmm28
; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm0 = ymm11[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm4 = [8,9,12,13,12,13,10,11,8,9,10,11,12,13,14,15]
; AVX512DQ-FCP-NEXT: vpshufb %xmm4, %xmm3, %xmm2
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm1 = [0,1,0,1,10,10,10,10]
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm1 = [0,0,0,1,0,10,10,0]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm1, %zmm2
; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX512DQ-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX512DQ-FCP-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11]
-; AVX512DQ-FCP-NEXT: vmovdqa (%rsp), %xmm3 # 16-byte Reload
+; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX512DQ-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX512DQ-FCP-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm24, %zmm3
; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512DQ-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512DQ-FCP-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # ymm6 = ymm6[0],mem[0],ymm6[1],mem[1],ymm6[2],mem[2],ymm6[3],mem[3],ymm6[8],mem[8],ymm6[9],mem[9],ymm6[10],mem[10],ymm6[11],mem[11]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm28, %zmm6
-; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm3, %zmm6 {%k1}
+; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # ymm7 = ymm7[0],mem[0],ymm7[1],mem[1],ymm7[2],mem[2],ymm7[3],mem[3],ymm7[8],mem[8],ymm7[9],mem[9],ymm7[10],mem[10],ymm7[11],mem[11]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm30, %zmm7
+; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm3, %zmm7 {%k1}
; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX512DQ-FCP-NEXT: vpshufb %xmm9, %xmm0, %xmm0
-; AVX512DQ-FCP-NEXT: vmovdqa %ymm6, %ymm3
-; AVX512DQ-FCP-NEXT: vpermt2d %ymm0, %ymm5, %ymm3
+; AVX512DQ-FCP-NEXT: vpshufb %xmm5, %xmm0, %xmm0
+; AVX512DQ-FCP-NEXT: vmovdqa %ymm7, %ymm3
+; AVX512DQ-FCP-NEXT: vpermt2d %ymm0, %ymm6, %ymm3
; AVX512DQ-FCP-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512DQ-FCP-NEXT: # ymm0 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm25, %zmm6
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm6, %zmm3, %zmm27
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm27, %zmm7
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm7, %zmm3, %zmm29
; AVX512DQ-FCP-NEXT: vpshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512DQ-FCP-NEXT: # ymm0 = mem[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX512DQ-FCP-NEXT: vpshufb %xmm4, %xmm3, %xmm6
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm1, %zmm6
+; AVX512DQ-FCP-NEXT: vpshufb %xmm4, %xmm3, %xmm7
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm1, %zmm7
; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX512DQ-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX512DQ-FCP-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11]
@@ -9991,71 +9985,71 @@ define void @store_i16_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX512DQ-FCP-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm0, %zmm24, %zmm3
-; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX512DQ-FCP-NEXT: vmovdqa (%rsp), %xmm0 # 16-byte Reload
; AVX512DQ-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512DQ-FCP-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX512DQ-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX512DQ-FCP-NEXT: # ymm10 = ymm10[0],mem[0],ymm10[1],mem[1],ymm10[2],mem[2],ymm10[3],mem[3],ymm10[8],mem[8],ymm10[9],mem[9],ymm10[10],mem[10],ymm10[11],mem[11]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm28, %zmm10
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm30, %zmm10
; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm3, %zmm10 {%k1}
; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX512DQ-FCP-NEXT: vpshufb %xmm9, %xmm0, %xmm0
+; AVX512DQ-FCP-NEXT: vpshufb %xmm5, %xmm0, %xmm0
; AVX512DQ-FCP-NEXT: vmovdqa %ymm10, %ymm3
-; AVX512DQ-FCP-NEXT: vpermt2d %ymm0, %ymm5, %ymm3
+; AVX512DQ-FCP-NEXT: vpermt2d %ymm0, %ymm6, %ymm3
; AVX512DQ-FCP-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512DQ-FCP-NEXT: # ymm0 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm25, %zmm10
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm27, %zmm10
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm10, %zmm3, %zmm3
; AVX512DQ-FCP-NEXT: vpshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
; AVX512DQ-FCP-NEXT: # ymm10 = mem[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512DQ-FCP-NEXT: vpshufb %xmm4, %xmm0, %xmm0
; AVX512DQ-FCP-NEXT: vpermt2q %zmm10, %zmm1, %zmm0
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm31, %ymm11
; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm10 = ymm10[0],ymm11[0],ymm10[1],ymm11[1],ymm10[2],ymm11[2],ymm10[3],ymm11[3],ymm10[8],ymm11[8],ymm10[9],ymm11[9],ymm10[10],ymm11[10],ymm10[11],ymm11[11]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # ymm10 = ymm10[0],mem[0],ymm10[1],mem[1],ymm10[2],mem[2],ymm10[3],mem[3],ymm10[8],mem[8],ymm10[9],mem[9],ymm10[10],mem[10],ymm10[11],mem[11]
; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm18, %xmm11
; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm17, %xmm12
; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm11 = xmm11[4],xmm12[4],xmm11[5],xmm12[5],xmm11[6],xmm12[6],xmm11[7],xmm12[7]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm10, %zmm24, %zmm11
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm20, %xmm10
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm21, %xmm10
; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm19, %xmm12
; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm10 = xmm10[4],xmm12[4],xmm10[5],xmm12[5],xmm10[6],xmm12[6],xmm10[7],xmm12[7]
; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX512DQ-FCP-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
; AVX512DQ-FCP-NEXT: # ymm12 = ymm12[0],mem[0],ymm12[1],mem[1],ymm12[2],mem[2],ymm12[3],mem[3],ymm12[8],mem[8],ymm12[9],mem[9],ymm12[10],mem[10],ymm12[11],mem[11]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm10, %zmm28, %zmm12
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm10, %zmm30, %zmm12
; AVX512DQ-FCP-NEXT: vmovdqa32 %zmm11, %zmm12 {%k1}
; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm16, %xmm10
-; AVX512DQ-FCP-NEXT: vpshufb %xmm9, %xmm10, %xmm9
-; AVX512DQ-FCP-NEXT: vpermi2d %ymm9, %ymm12, %ymm5
-; AVX512DQ-FCP-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # ymm9 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpermt2d %zmm9, %zmm25, %zmm12
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm12, %zmm5, %zmm5
+; AVX512DQ-FCP-NEXT: vpshufb %xmm5, %xmm10, %xmm5
+; AVX512DQ-FCP-NEXT: vpermi2d %ymm5, %ymm12, %ymm6
+; AVX512DQ-FCP-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # ymm5 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
+; AVX512DQ-FCP-NEXT: vpermt2d %zmm5, %zmm27, %zmm12
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm12, %zmm6, %zmm5
; AVX512DQ-FCP-NEXT: vpshufb %xmm4, %xmm15, %xmm4
-; AVX512DQ-FCP-NEXT: vpshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # ymm9 = mem[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm9, %zmm1, %zmm4
+; AVX512DQ-FCP-NEXT: vpshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # ymm6 = mem[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm1, %zmm4
; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm4 ^ (zmm1 & (zmm4 ^ zmm5))
; AVX512DQ-FCP-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm4, 256(%rax)
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (zmm1 & (zmm0 ^ zmm3))
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm0, 448(%rax)
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm6 ^ (zmm1 & (zmm6 ^ zmm27))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm6, 640(%rax)
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm1 & (zmm2 ^ zmm26))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (zmm1 & (zmm7 ^ zmm29))
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm7, 640(%rax)
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm1 & (zmm2 ^ zmm28))
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, 64(%rax)
; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (zmm0 & (zmm7 ^ zmm23))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm7, (%rax)
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm8 ^ (zmm0 & (zmm8 ^ zmm22))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm8, 192(%rax)
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm8 ^ (zmm0 & (zmm8 ^ zmm23))
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm8, (%rax)
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm9 = zmm9 ^ (zmm0 & (zmm9 ^ zmm22))
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm9, 192(%rax)
; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,0]
; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm1 & (zmm2 ^ zmm21))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm1 & (zmm2 ^ zmm25))
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, 128(%rax)
; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm2 # 64-byte Folded Reload
@@ -10066,18 +10060,16 @@ define void @store_i16_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: # zmm2 = zmm2 ^ (zmm0 & (zmm2 ^ mem))
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, 384(%rax)
; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
-; AVX512DQ-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 # 64-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # zmm2 = zmm2 ^ (zmm0 & (zmm2 ^ mem))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, 576(%rax)
-; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512DQ-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # zmm0 = zmm0 ^ (zmm1 & (zmm0 ^ mem))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm0, 512(%rax)
-; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512DQ-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # zmm0 = zmm0 ^ (zmm1 & (zmm0 ^ mem))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm0, 704(%rax)
-; AVX512DQ-FCP-NEXT: addq $1320, %rsp # imm = 0x528
+; AVX512DQ-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm2 # 64-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # zmm2 = zmm2 ^ (zmm1 & (zmm2 ^ mem))
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, 512(%rax)
+; AVX512DQ-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm20 # 64-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # zmm20 = zmm20 ^ (zmm1 & (zmm20 ^ mem))
+; AVX512DQ-FCP-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm26 # 64-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # zmm26 = zmm26 ^ (zmm0 & (zmm26 ^ mem))
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm26, 576(%rax)
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm20, 704(%rax)
+; AVX512DQ-FCP-NEXT: addq $1224, %rsp # imm = 0x4C8
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
index fdbe919c0f73e..6f50d61f4d1f4 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
@@ -6242,7 +6242,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT: vmovdqa 32(%r9), %ymm1
; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm1, %ymm0
-; AVX512-FCP-NEXT: vmovdqa64 %ymm1, %ymm20
+; AVX512-FCP-NEXT: vmovdqa64 %ymm1, %ymm22
; AVX512-FCP-NEXT: vmovdqa 32(%r8), %ymm2
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm2, %ymm1
; AVX512-FCP-NEXT: vmovdqa64 %ymm2, %ymm16
@@ -6331,7 +6331,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm31 = [2,2,3,3,10,0,11,10]
; AVX512-FCP-NEXT: vpermi2q %zmm3, %zmm4, %zmm31
; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm0, %xmm22
+; AVX512-FCP-NEXT: vmovdqa64 %xmm0, %xmm20
; AVX512-FCP-NEXT: vmovdqa (%rdi), %xmm4
; AVX512-FCP-NEXT: vmovdqa (%rsi), %xmm3
; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
@@ -6352,7 +6352,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm7[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm9 = ymm14[0,1,1,3,4,5,5,7]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm8 = ymm9[0,1],ymm1[2],ymm9[3,4],ymm1[5],ymm9[6,7,8,9],ymm1[10],ymm9[11,12],ymm1[13],ymm9[14,15]
-; AVX512-FCP-NEXT: vmovdqa64 %ymm20, %ymm13
+; AVX512-FCP-NEXT: vmovdqa64 %ymm22, %ymm13
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm13, %ymm1
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm6 = ymm16[0,0,2,1,4,4,6,5]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm6[3],ymm1[4,5],ymm6[6],ymm1[7,8,9,10],ymm6[11],ymm1[12,13],ymm6[14],ymm1[15]
@@ -6361,9 +6361,8 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm9
; AVX512-FCP-NEXT: vmovdqa 32(%rax), %ymm1
; AVX512-FCP-NEXT: vpshufb %ymm2, %ymm1, %ymm2
-; AVX512-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm20 = [0,5,4,0,0,6,5,0,0,5,4,0,0,6,5,0]
-; AVX512-FCP-NEXT: # zmm20 = mem[0,1,2,3,0,1,2,3]
-; AVX512-FCP-NEXT: vpermd %ymm1, %ymm20, %ymm1
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm22 = [0,0,4,0,0,0,5,0,0,5,0,0,0,6,0,0]
+; AVX512-FCP-NEXT: vpermd %ymm1, %ymm22, %ymm1
; AVX512-FCP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm2 = ymm7[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
@@ -6413,8 +6412,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm12 = zmm24 ^ (zmm17 & (zmm12 ^ zmm24))
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm14 = zmm14 ^ (mem & (zmm14 ^ zmm12))
; AVX512-FCP-NEXT: vmovdqa64 (%rax), %zmm5
-; AVX512-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm6 = [14,13,8,8,15,14,8,15,14,13,8,8,15,14,8,15]
-; AVX512-FCP-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3]
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm6 = [0,13,0,0,0,14,0,0,14,0,0,0,15,0,0,15]
; AVX512-FCP-NEXT: vpermd %zmm5, %zmm6, %zmm6
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm6 ^ (mem & (zmm6 ^ zmm14))
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm12 = [6,0,0,0,7,0,0,7]
@@ -6432,7 +6430,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm24 # 32-byte Reload
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm11 = ymm24[2,2,2,2,6,6,6,6]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm7 = ymm11[0],ymm7[1],ymm11[2,3],ymm7[4],ymm11[5,6,7,8],ymm7[9],ymm11[10,11],ymm7[12],ymm11[13,14,15]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm22, %xmm2
+; AVX512-FCP-NEXT: vmovdqa64 %xmm20, %xmm2
; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm11 = xmm2[0,2,3,3,4,5,6,7]
; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,0,2,1]
; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm14 = ymm19[0,0,1,3]
@@ -6444,7 +6442,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3,4],ymm2[5],ymm0[6,7,8,9],ymm2[10],ymm0[11,12],ymm2[13],ymm0[14,15]
; AVX512-FCP-NEXT: vpermt2q %zmm7, %zmm16, %zmm0
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm3))
-; AVX512-FCP-NEXT: vpermd %zmm5, %zmm20, %zmm2
+; AVX512-FCP-NEXT: vpermd %zmm5, %zmm22, %zmm2
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm31))
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm0))
; AVX512-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm11, %zmm0 # 32-byte Folded Reload
@@ -6867,7 +6865,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa 32(%r9), %ymm1
; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm1, %ymm0
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm1, %ymm20
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm1, %ymm22
; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %ymm2
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm2, %ymm1
; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm2, %ymm16
@@ -6956,7 +6954,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm31 = [2,2,3,3,10,0,11,10]
; AVX512DQ-FCP-NEXT: vpermi2q %zmm3, %zmm4, %zmm31
; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm0, %xmm22
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm0, %xmm20
; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %xmm4
; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %xmm3
; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
@@ -6977,7 +6975,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm7[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm9 = ymm14[0,1,1,3,4,5,5,7]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm8 = ymm9[0,1],ymm1[2],ymm9[3,4],ymm1[5],ymm9[6,7,8,9],ymm1[10],ymm9[11,12],ymm1[13],ymm9[14,15]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm20, %ymm13
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm22, %ymm13
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm13, %ymm1
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm6 = ymm16[0,0,2,1,4,4,6,5]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm6[3],ymm1[4,5],ymm6[6],ymm1[7,8,9,10],ymm6[11],ymm1[12,13],ymm6[14],ymm1[15]
@@ -6986,9 +6984,8 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vpermi2q %zmm1, %zmm0, %zmm9
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rax), %ymm1
; AVX512DQ-FCP-NEXT: vpshufb %ymm2, %ymm1, %ymm2
-; AVX512DQ-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm20 = [0,5,4,0,0,6,5,0,0,5,4,0,0,6,5,0]
-; AVX512DQ-FCP-NEXT: # zmm20 = mem[0,1,2,3,0,1,2,3]
-; AVX512DQ-FCP-NEXT: vpermd %ymm1, %ymm20, %ymm1
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm22 = [0,0,4,0,0,0,5,0,0,5,0,0,0,6,0,0]
+; AVX512DQ-FCP-NEXT: vpermd %ymm1, %ymm22, %ymm1
; AVX512DQ-FCP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm2 = ymm7[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
@@ -7038,8 +7035,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm12 = zmm24 ^ (zmm17 & (zmm12 ^ zmm24))
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm14 = zmm14 ^ (mem & (zmm14 ^ zmm12))
; AVX512DQ-FCP-NEXT: vmovdqa64 (%rax), %zmm5
-; AVX512DQ-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm6 = [14,13,8,8,15,14,8,15,14,13,8,8,15,14,8,15]
-; AVX512DQ-FCP-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3]
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm6 = [0,13,0,0,0,14,0,0,14,0,0,0,15,0,0,15]
; AVX512DQ-FCP-NEXT: vpermd %zmm5, %zmm6, %zmm6
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm6 ^ (mem & (zmm6 ^ zmm14))
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm12 = [6,0,0,0,7,0,0,7]
@@ -7057,7 +7053,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm24 # 32-byte Reload
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm11 = ymm24[2,2,2,2,6,6,6,6]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm7 = ymm11[0],ymm7[1],ymm11[2,3],ymm7[4],ymm11[5,6,7,8],ymm7[9],ymm11[10,11],ymm7[12],ymm11[13,14,15]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm22, %xmm2
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm20, %xmm2
; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm11 = xmm2[0,2,3,3,4,5,6,7]
; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,0,2,1]
; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm14 = ymm19[0,0,1,3]
@@ -7069,7 +7065,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3,4],ymm2[5],ymm0[6,7,8,9],ymm2[10],ymm0[11,12],ymm2[13],ymm0[14,15]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm7, %zmm16, %zmm0
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm3))
-; AVX512DQ-FCP-NEXT: vpermd %zmm5, %zmm20, %zmm2
+; AVX512DQ-FCP-NEXT: vpermd %zmm5, %zmm22, %zmm2
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm31))
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm0))
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm11, %zmm0 # 32-byte Folded Reload
@@ -13229,7 +13225,7 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX512-FCP-LABEL: store_i16_stride7_vf64:
; AVX512-FCP: # %bb.0:
-; AVX512-FCP-NEXT: subq $1544, %rsp # imm = 0x608
+; AVX512-FCP-NEXT: subq $1480, %rsp # imm = 0x5C8
; AVX512-FCP-NEXT: vmovdqa 96(%rcx), %ymm3
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128]
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm3, %ymm0
@@ -13256,48 +13252,48 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vmovdqa 64(%rcx), %ymm0
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm0, %ymm14
-; AVX512-FCP-NEXT: vmovdqa64 %ymm0, %ymm29
+; AVX512-FCP-NEXT: vmovdqa64 %ymm0, %ymm27
; AVX512-FCP-NEXT: vmovdqa 64(%rdx), %ymm0
; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm0, %ymm15
-; AVX512-FCP-NEXT: vmovdqa64 %ymm0, %ymm18
-; AVX512-FCP-NEXT: vpor %ymm14, %ymm15, %ymm14
-; AVX512-FCP-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vmovdqa64 %ymm0, %ymm29
+; AVX512-FCP-NEXT: vpor %ymm14, %ymm15, %ymm13
+; AVX512-FCP-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vmovdqa 64(%rsi), %ymm0
; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm0, %ymm14
-; AVX512-FCP-NEXT: vmovdqa64 %ymm0, %ymm27
+; AVX512-FCP-NEXT: vmovdqa64 %ymm0, %ymm22
; AVX512-FCP-NEXT: vmovdqa 64(%rdi), %ymm0
; AVX512-FCP-NEXT: vpshufb %ymm11, %ymm0, %ymm15
; AVX512-FCP-NEXT: vmovdqa64 %ymm0, %ymm28
-; AVX512-FCP-NEXT: vpor %ymm14, %ymm15, %ymm14
-; AVX512-FCP-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpor %ymm14, %ymm15, %ymm13
+; AVX512-FCP-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vmovdqa (%r9), %ymm0
; AVX512-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm0, %ymm14
; AVX512-FCP-NEXT: vmovdqa (%r8), %ymm0
; AVX512-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vpshufb %ymm12, %ymm0, %ymm15
-; AVX512-FCP-NEXT: vpor %ymm14, %ymm15, %ymm14
-; AVX512-FCP-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpor %ymm14, %ymm15, %ymm13
+; AVX512-FCP-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vmovdqa (%rcx), %ymm0
; AVX512-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm0, %ymm14
; AVX512-FCP-NEXT: vmovdqa (%rdx), %ymm0
; AVX512-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm0, %ymm15
-; AVX512-FCP-NEXT: vpor %ymm14, %ymm15, %ymm14
-; AVX512-FCP-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpor %ymm14, %ymm15, %ymm13
+; AVX512-FCP-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vmovdqa (%rsi), %ymm0
; AVX512-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm0, %ymm14
+; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm0, %ymm15
; AVX512-FCP-NEXT: vmovdqa (%rdi), %ymm0
; AVX512-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshufb %ymm11, %ymm0, %ymm15
-; AVX512-FCP-NEXT: vpor %ymm14, %ymm15, %ymm14
-; AVX512-FCP-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpshufb %ymm11, %ymm0, %ymm13
+; AVX512-FCP-NEXT: vpor %ymm15, %ymm13, %ymm13
+; AVX512-FCP-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vmovdqa 32(%rcx), %ymm15
; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm15, %ymm6
-; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %ymm13
-; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm13, %ymm8
+; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %ymm14
+; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm14, %ymm8
; AVX512-FCP-NEXT: vpor %ymm6, %ymm8, %ymm6
; AVX512-FCP-NEXT: vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT: vmovdqa 32(%rsi), %ymm0
@@ -13315,32 +13311,32 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm12 = ymm1[0,1,1,3,4,5,5,7]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm11 = ymm12[0,1],ymm11[2],ymm12[3,4],ymm11[5],ymm12[6,7,8,9],ymm11[10],ymm12[11,12],ymm11[13],ymm12[14,15]
; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm11 = ymm11[2,1,3,2]
-; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm14 = [18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
-; AVX512-FCP-NEXT: # ymm14 = mem[0,1,0,1]
-; AVX512-FCP-NEXT: vpshufb %ymm14, %ymm2, %ymm12
-; AVX512-FCP-NEXT: vmovdqa64 %ymm14, %ymm17
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm14 = ymm7[1,1,1,1,5,5,5,5]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm12 = ymm12[0,1],ymm14[2],ymm12[3,4],ymm14[5],ymm12[6,7,8,9],ymm14[10],ymm12[11,12],ymm14[13],ymm12[14,15]
+; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm13 = [18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
+; AVX512-FCP-NEXT: # ymm13 = mem[0,1,0,1]
+; AVX512-FCP-NEXT: vpshufb %ymm13, %ymm2, %ymm12
+; AVX512-FCP-NEXT: vmovdqa64 %ymm13, %ymm17
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm13 = ymm7[1,1,1,1,5,5,5,5]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm12 = ymm12[0,1],ymm13[2],ymm12[3,4],ymm13[5],ymm12[6,7,8,9],ymm13[10],ymm12[11,12],ymm13[13],ymm12[14,15]
; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm12 = ymm12[2,2,2,3]
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm11, %zmm16, %zmm11
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm12, %zmm5, %zmm12
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm12 = zmm11 ^ (mem & (zmm12 ^ zmm11))
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm14 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535]
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm13 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535]
; AVX512-FCP-NEXT: vmovdqa 96(%r8), %ymm5
; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm11 = ymm5[u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm5[14,15,u,u,u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm5[16,17,u,u,u,u],zero,zero
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} ymm11 = ymm11 | (ymm12 & ymm14)
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} ymm11 = ymm11 | (ymm12 & ymm13)
; AVX512-FCP-NEXT: vmovdqa 96(%r9), %ymm6
-; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm6, %ymm4
; AVX512-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm6, %ymm4
; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm16 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} ymm4 = ymm4 | (ymm11 & ymm16)
-; AVX512-FCP-NEXT: vmovdqa64 %zmm16, %zmm30
+; AVX512-FCP-NEXT: vmovdqa64 %zmm16, %zmm20
; AVX512-FCP-NEXT: vextracti64x4 $1, %zmm12, %ymm11
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm12 = [0,4,0,0,0,5,0,0]
; AVX512-FCP-NEXT: vpermd %ymm5, %ymm12, %ymm12
; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm16 = [65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0]
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} ymm12 = ymm12 ^ (ymm16 & (ymm12 ^ ymm11))
-; AVX512-FCP-NEXT: vmovdqa64 %zmm16, %zmm20
+; AVX512-FCP-NEXT: vmovdqa64 %zmm16, %zmm18
; AVX512-FCP-NEXT: vprold $16, %ymm6, %ymm11
; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm11 = ymm11[2,2,2,2]
; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm16 = [65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535]
@@ -13351,19 +13347,19 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm11 = [0,1,4,5,4,5,5,7]
; AVX512-FCP-NEXT: vmovdqa 96(%rax), %ymm4
; AVX512-FCP-NEXT: vpermd %ymm4, %ymm11, %ymm11
-; AVX512-FCP-NEXT: vpandn %ymm11, %ymm14, %ymm11
-; AVX512-FCP-NEXT: vmovdqa64 %zmm14, %zmm22
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm12 = [12,13,128,128,128,128,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128]
-; AVX512-FCP-NEXT: vpshufb %ymm12, %ymm4, %ymm4
+; AVX512-FCP-NEXT: vpandn %ymm11, %ymm13, %ymm11
+; AVX512-FCP-NEXT: vmovdqa64 %zmm13, %zmm12
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm13 = [12,13,128,128,128,128,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128]
+; AVX512-FCP-NEXT: vpshufb %ymm13, %ymm4, %ymm4
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm11, %zmm4, %zmm4
; AVX512-FCP-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm14 = [u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29]
-; AVX512-FCP-NEXT: vpshufb %ymm14, %ymm2, %ymm4
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm6 = [u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29]
+; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm2, %ymm4
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm11 = ymm7[3,3,3,3,7,7,7,7]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm11[0,1,2],ymm4[3],ymm11[4,5],ymm4[6],ymm11[7,8,9,10],ymm4[11],ymm11[12,13],ymm4[14],ymm11[15]
-; AVX512-FCP-NEXT: vpmovsxdq {{.*#+}} ymm6 = [151522058,0,421010202,421010202]
-; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm2, %ymm2
-; AVX512-FCP-NEXT: vmovdqa64 %ymm6, %ymm23
+; AVX512-FCP-NEXT: vpmovsxdq {{.*#+}} ymm11 = [151522058,0,421010202,421010202]
+; AVX512-FCP-NEXT: vpshufb %ymm11, %ymm2, %ymm2
+; AVX512-FCP-NEXT: vmovdqa64 %ymm11, %ymm24
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm7 = ymm7[2,2,2,2,6,6,6,6]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm7[0],ymm2[1],ymm7[2,3],ymm2[4],ymm7[5,6,7,8],ymm2[9],ymm7[10,11],ymm2[12],ymm7[13,14,15]
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm7 = [0,2,2,3,10,9,11,11]
@@ -13373,16 +13369,15 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vmovdqa64 %ymm11, %ymm16
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm11 = ymm1[3,3,3,3,7,7,7,7]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm11[0],ymm4[1],ymm11[2,3],ymm4[4],ymm11[5,6,7,8],ymm4[9],ymm11[10,11],ymm4[12],ymm11[13,14,15]
-; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm6 = [22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27,22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27]
-; AVX512-FCP-NEXT: # ymm6 = mem[0,1,0,1]
-; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX512-FCP-NEXT: vmovdqa64 %ymm6, %ymm24
+; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm11 = [22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27,22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27]
+; AVX512-FCP-NEXT: # ymm11 = mem[0,1,0,1]
+; AVX512-FCP-NEXT: vpshufb %ymm11, %ymm3, %ymm3
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0,1],ymm1[2],ymm3[3,4],ymm1[5],ymm3[6,7,8,9],ymm1[10],ymm3[11,12],ymm1[13],ymm3[14,15]
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm19 = [2,2,2,3,8,10,10,11]
; AVX512-FCP-NEXT: vpermt2q %zmm4, %zmm19, %zmm1
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm11 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm2 ^ (zmm11 & (zmm1 ^ zmm2))
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm23 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm2 ^ (zmm23 & (zmm1 ^ zmm2))
; AVX512-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm1 = [5,0,0,0,6,0,0,6]
; AVX512-FCP-NEXT: vpermd %ymm5, %ymm1, %ymm1
@@ -13392,45 +13387,45 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vpandn %ymm1, %ymm3, %ymm1
; AVX512-FCP-NEXT: vmovdqa 64(%rax), %ymm2
; AVX512-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshufb %ymm12, %ymm2, %ymm2
+; AVX512-FCP-NEXT: vpshufb %ymm13, %ymm2, %ymm2
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
; AVX512-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT: vpbroadcastd 8(%rax), %ymm1
; AVX512-FCP-NEXT: vpandn %ymm1, %ymm3, %ymm1
; AVX512-FCP-NEXT: vmovdqa (%rax), %ymm2
; AVX512-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshufb %ymm12, %ymm2, %ymm2
+; AVX512-FCP-NEXT: vpshufb %ymm13, %ymm2, %ymm2
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
; AVX512-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vpbroadcastd {{.*#+}} ymm6 = [18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
-; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm10, %ymm1
+; AVX512-FCP-NEXT: vpbroadcastd {{.*#+}} ymm5 = [18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
+; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm10, %ymm1
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm2 = ymm8[0,0,2,1,4,4,6,5]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7,8,9,10],ymm2[11],ymm1[12,13],ymm2[14],ymm1[15]
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm2 = [0,1,0,3,10,10,11,11]
; AVX512-FCP-NEXT: vpermi2q %zmm1, %zmm21, %zmm2
; AVX512-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshufb %ymm14, %ymm0, %ymm1
+; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm0, %ymm1
; AVX512-FCP-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5],ymm1[6],ymm2[7,8,9,10],ymm1[11],ymm2[12,13],ymm1[14],ymm2[15]
-; AVX512-FCP-NEXT: vmovdqa64 %ymm23, %ymm5
-; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm0, %ymm2
+; AVX512-FCP-NEXT: vmovdqa64 %ymm24, %ymm13
+; AVX512-FCP-NEXT: vpshufb %ymm13, %ymm0, %ymm2
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm9[2,2,2,2,6,6,6,6]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5,6,7,8],ymm2[9],ymm3[10,11],ymm2[12],ymm3[13,14,15]
; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm7, %zmm2
; AVX512-FCP-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vmovdqa64 %ymm16, %ymm7
-; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm15, %ymm1
-; AVX512-FCP-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm13[3,3,3,3,7,7,7,7]
+; AVX512-FCP-NEXT: vmovdqa64 %ymm16, %ymm6
+; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm15, %ymm1
+; AVX512-FCP-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm14[3,3,3,3,7,7,7,7]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5,6,7,8],ymm1[9],ymm3[10,11],ymm1[12],ymm3[13,14,15]
-; AVX512-FCP-NEXT: vmovdqa64 %ymm24, %ymm12
-; AVX512-FCP-NEXT: vpshufb %ymm12, %ymm15, %ymm3
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm13[2,2,2,2,6,6,6,6]
+; AVX512-FCP-NEXT: vmovdqa %ymm11, %ymm7
+; AVX512-FCP-NEXT: vpshufb %ymm11, %ymm15, %ymm3
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm14[2,2,2,2,6,6,6,6]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7,8,9],ymm4[10],ymm3[11,12],ymm4[13],ymm3[14,15]
; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm19, %zmm0
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm2 ^ (zmm11 & (zmm0 ^ zmm2))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm2 ^ (zmm23 & (zmm0 ^ zmm2))
; AVX512-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm1 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm2 = ymm8[3,3,3,3,7,7,7,7]
@@ -13438,348 +13433,343 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vprold $16, %ymm10, %ymm2
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm8[1,2,2,3,5,6,6,7]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7,8,9],ymm2[10],ymm3[11,12],ymm2[13],ymm3[14,15]
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm13 = [2,1,3,2,10,10,10,11]
-; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm13, %zmm2
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm11 = [2,1,3,2,10,10,10,11]
+; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm11, %zmm2
; AVX512-FCP-NEXT: vmovdqa64 (%rax), %zmm16
-; AVX512-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [14,13,8,8,15,14,8,15,14,13,8,8,15,14,8,15]
-; AVX512-FCP-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3]
-; AVX512-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm0 = [0,13,0,0,0,14,0,0,14,0,0,0,15,0,0,15]
; AVX512-FCP-NEXT: vpermd %zmm16, %zmm0, %zmm0
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (zmm20 & (zmm0 ^ zmm2))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (zmm18 & (zmm0 ^ zmm2))
; AVX512-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vmovdqa64 %ymm27, %ymm0
-; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm0, %ymm1
-; AVX512-FCP-NEXT: vmovdqa64 %ymm23, %ymm21
+; AVX512-FCP-NEXT: vmovdqa64 %ymm22, %ymm15
+; AVX512-FCP-NEXT: vpshufb %ymm13, %ymm15, %ymm1
+; AVX512-FCP-NEXT: vmovdqa64 %ymm24, %ymm21
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm2 = ymm28[2,2,2,2,6,6,6,6]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6,7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13,14,15]
; AVX512-FCP-NEXT: vmovdqa64 %ymm17, %ymm9
-; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm0, %ymm2
+; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm15, %ymm2
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm28[1,1,1,1,5,5,5,5]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm3 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7,8,9],ymm3[10],ymm2[11,12],ymm3[13],ymm2[14,15]
-; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm19, %zmm3
-; AVX512-FCP-NEXT: vmovdqa64 %ymm29, %ymm8
-; AVX512-FCP-NEXT: vpshufb %ymm12, %ymm8, %ymm1
-; AVX512-FCP-NEXT: vmovdqa64 %ymm24, %ymm10
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm2 = ymm18[2,2,2,2,6,6,6,6]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm8[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm18[0,1,1,3,4,5,5,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0,1],ymm2[2],ymm4[3,4],ymm2[5],ymm4[6,7,8,9],ymm2[10],ymm4[11,12],ymm2[13],ymm4[14,15]
-; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm13, %zmm5
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm3 ^ (mem & (zmm5 ^ zmm3))
-; AVX512-FCP-NEXT: vmovdqa64 %ymm25, %ymm0
-; AVX512-FCP-NEXT: vmovdqu64 %ymm25, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512-FCP-NEXT: vprold $16, %ymm25, %ymm1
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7,8,9],ymm3[10],ymm2[11,12],ymm3[13],ymm2[14,15]
+; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm19, %zmm2
+; AVX512-FCP-NEXT: vmovdqa64 %ymm27, %ymm0
+; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm0, %ymm1
+; AVX512-FCP-NEXT: vmovdqa %ymm7, %ymm10
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm29[2,2,2,2,6,6,6,6]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm0[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
+; AVX512-FCP-NEXT: vmovdqa64 %ymm27, %ymm7
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm29[0,1,1,3,4,5,5,7]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7,8,9],ymm3[10],ymm4[11,12],ymm3[13],ymm4[14,15]
+; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm11, %zmm4
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm2 ^ (zmm1 & (zmm4 ^ zmm2))
+; AVX512-FCP-NEXT: vmovdqa64 %ymm25, %ymm13
+; AVX512-FCP-NEXT: vprold $16, %ymm25, %ymm2
; AVX512-FCP-NEXT: vmovdqa64 %ymm26, %ymm25
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm26[1,2,2,3,5,6,6,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0,1],ymm1[2],ymm3[3,4],ymm1[5],ymm3[6,7,8,9],ymm1[10],ymm3[11,12],ymm1[13],ymm3[14,15]
-; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm0, %ymm3
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm26[0,0,2,1,4,4,6,5]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3],ymm3[4,5],ymm4[6],ymm3[7,8,9,10],ymm4[11],ymm3[12,13],ymm4[14],ymm3[15]
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm0 = [2,2,3,3,10,9,11,10]
-; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm0, %zmm3
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7,8,9],ymm2[10],ymm3[11,12],ymm2[13],ymm3[14,15]
+; AVX512-FCP-NEXT: vpshufb %ymm5, %ymm13, %ymm3
+; AVX512-FCP-NEXT: vmovdqa64 %ymm13, %ymm18
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm26[0,0,2,1,4,4,6,5]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3],ymm3[4,5],ymm5[6],ymm3[7,8,9,10],ymm5[11],ymm3[12,13],ymm5[14],ymm3[15]
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm27 = [2,2,3,3,10,9,11,10]
+; AVX512-FCP-NEXT: vpermt2q %zmm2, %zmm27, %zmm3
; AVX512-FCP-NEXT: vmovdqa64 64(%rax), %zmm0
; AVX512-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm23 = [0,5,4,0,0,6,5,0,0,5,4,0,0,6,5,0]
-; AVX512-FCP-NEXT: # zmm23 = mem[0,1,2,3,0,1,2,3]
-; AVX512-FCP-NEXT: vpermd %zmm0, %zmm23, %zmm0
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (zmm22 & (zmm0 ^ zmm3))
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm5))
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm2 = [0,0,4,0,0,0,5,0,0,5,0,0,0,6,0,0]
+; AVX512-FCP-NEXT: vpermd %zmm0, %zmm2, %zmm0
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (zmm12 & (zmm0 ^ zmm3))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm4))
; AVX512-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm8, %ymm1
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm0 = ymm18[3,3,3,3,7,7,7,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm12 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6,7,8],ymm1[9],ymm0[10,11],ymm1[12],ymm0[13,14,15]
-; AVX512-FCP-NEXT: vmovdqa 64(%rcx), %xmm5
-; AVX512-FCP-NEXT: vmovdqa 64(%rdx), %xmm6
-; AVX512-FCP-NEXT: vpbroadcastq {{.*#+}} xmm4 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
-; AVX512-FCP-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpshufb %xmm4, %xmm5, %xmm1
-; AVX512-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm3 = xmm6[1,1,2,2]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2,3],xmm1[4],xmm3[5,6],xmm1[7]
+; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm7, %ymm2
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm0 = ymm29[3,3,3,3,7,7,7,7]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm13 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6,7,8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14,15]
+; AVX512-FCP-NEXT: vmovdqa 64(%rcx), %xmm3
+; AVX512-FCP-NEXT: vmovdqa 64(%rdx), %xmm4
+; AVX512-FCP-NEXT: vpbroadcastq {{.*#+}} xmm12 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
+; AVX512-FCP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vpshufb %xmm12, %xmm3, %xmm2
+; AVX512-FCP-NEXT: vmovdqa %xmm4, %xmm5
+; AVX512-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[1,1,2,2]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0],xmm2[1],xmm4[2,3],xmm2[4],xmm4[5,6],xmm2[7]
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm14 = [0,1,2,3,0,1,2,3,6,7,4,5,6,7,4,5]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; AVX512-FCP-NEXT: vpshufb %xmm14, %xmm3, %xmm3
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm22 = [8,9,9,0,0,0,1,1]
-; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm22, %zmm1
-; AVX512-FCP-NEXT: vmovdqa 64(%rdi), %xmm6
-; AVX512-FCP-NEXT: vmovdqa %xmm6, (%rsp) # 16-byte Spill
-; AVX512-FCP-NEXT: vmovdqa 64(%rsi), %xmm7
-; AVX512-FCP-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vprold $16, %xmm7, %xmm3
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm5 = xmm6[1,1,2,3]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm3 = xmm5[0,1],xmm3[2],xmm5[3,4],xmm3[5],xmm5[6,7]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; AVX512-FCP-NEXT: vpshufb %xmm14, %xmm4, %xmm4
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm24 = [8,9,9,0,0,0,1,1]
+; AVX512-FCP-NEXT: vpermt2q %zmm4, %zmm24, %zmm2
+; AVX512-FCP-NEXT: vmovdqa 64(%rdi), %xmm3
+; AVX512-FCP-NEXT: vmovdqa %xmm3, (%rsp) # 16-byte Spill
+; AVX512-FCP-NEXT: vmovdqa 64(%rsi), %xmm6
+; AVX512-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vprold $16, %xmm6, %xmm4
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm5 = xmm3[1,1,2,3]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1],xmm4[2],xmm5[3,4],xmm4[5],xmm5[6,7]
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm15 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3]
; AVX512-FCP-NEXT: vpshufb %xmm15, %xmm5, %xmm5
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm26 = [0,0,1,1,8,8,0,9]
-; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm26, %zmm5
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm1 ^ (zmm3 & (zmm5 ^ zmm1))
-; AVX512-FCP-NEXT: vmovdqa 64(%r9), %xmm1
+; AVX512-FCP-NEXT: vpermt2q %zmm4, %zmm26, %zmm5
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm4 = [65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm2 ^ (zmm4 & (zmm5 ^ zmm2))
+; AVX512-FCP-NEXT: vmovdqa 64(%r9), %xmm2
; AVX512-FCP-NEXT: vmovdqa 64(%r8), %xmm6
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm6[4],xmm1[4],xmm6[5],xmm1[5],xmm6[6],xmm1[6],xmm6[7],xmm1[7]
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm6[4],xmm2[4],xmm6[5],xmm2[5],xmm6[6],xmm2[6],xmm6[7],xmm2[7]
; AVX512-FCP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm6[0],xmm1[0],xmm6[1],xmm1[1],xmm6[2],xmm1[2],xmm6[3],xmm1[3]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3]
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm7 = [0,1,2,3,6,7,4,5,6,7,4,5,12,13,14,15]
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
-; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm1, %xmm6
-; AVX512-FCP-NEXT: vpshufb %xmm7, %xmm1, %xmm1
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm18 = [0,0,0,1,8,9,9,0]
-; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm18, %zmm1
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
+; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm2, %xmm6
+; AVX512-FCP-NEXT: vpshufb %xmm7, %xmm2, %xmm2
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm29 = [0,0,0,1,8,9,9,0]
+; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm29, %zmm2
; AVX512-FCP-NEXT: vpbroadcastd 64(%rax), %ymm6
; AVX512-FCP-NEXT: vpbroadcastd 68(%rax), %ymm8
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm6, %zmm29
-; AVX512-FCP-NEXT: vmovdqa64 %zmm30, %zmm11
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm29 = zmm29 ^ (zmm30 & (zmm29 ^ zmm1))
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm29 = zmm29 ^ (zmm1 & (zmm29 ^ zmm5))
-; AVX512-FCP-NEXT: vmovdqa (%rcx), %xmm8
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm6, %zmm30
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm30 = zmm30 ^ (zmm20 & (zmm30 ^ zmm2))
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm30 = zmm30 ^ (zmm2 & (zmm30 ^ zmm5))
+; AVX512-FCP-NEXT: vmovdqa (%rcx), %xmm3
+; AVX512-FCP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vmovdqa (%rdx), %xmm8
; AVX512-FCP-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vmovdqa (%rdx), %xmm0
-; AVX512-FCP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpshufb %xmm4, %xmm8, %xmm5
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm6 = xmm0[1,1,2,2]
+; AVX512-FCP-NEXT: vpshufb %xmm12, %xmm3, %xmm5
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm6 = xmm8[1,1,2,2]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2,3],xmm5[4],xmm6[5,6],xmm5[7]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm3[0],xmm8[1],xmm3[1],xmm8[2],xmm3[2],xmm8[3],xmm3[3]
; AVX512-FCP-NEXT: vpshufb %xmm14, %xmm6, %xmm6
-; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm22, %zmm5
-; AVX512-FCP-NEXT: vmovdqa (%rdi), %xmm0
-; AVX512-FCP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vmovdqa (%rsi), %xmm4
-; AVX512-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vprold $16, %xmm4, %xmm6
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm8 = xmm0[1,1,2,3]
+; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm24, %zmm5
+; AVX512-FCP-NEXT: vmovdqa (%rdi), %xmm3
+; AVX512-FCP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vmovdqa (%rsi), %xmm12
+; AVX512-FCP-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vprold $16, %xmm12, %xmm6
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm8 = xmm3[1,1,2,3]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm6 = xmm8[0,1],xmm6[2],xmm8[3,4],xmm6[5],xmm8[6,7]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm3[0],xmm12[0],xmm3[1],xmm12[1],xmm3[2],xmm12[2],xmm3[3],xmm12[3]
; AVX512-FCP-NEXT: vpshufb %xmm15, %xmm8, %xmm8
; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm26, %zmm8
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm5 ^ (zmm3 & (zmm8 ^ zmm5))
-; AVX512-FCP-NEXT: vmovdqa (%r9), %xmm3
+; AVX512-FCP-NEXT: vmovdqa64 %zmm26, %zmm24
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm5 ^ (zmm4 & (zmm8 ^ zmm5))
+; AVX512-FCP-NEXT: vmovdqa (%r9), %xmm4
; AVX512-FCP-NEXT: vmovdqa (%r8), %xmm5
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
-; AVX512-FCP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
-; AVX512-FCP-NEXT: vpshufb %xmm2, %xmm3, %xmm5
-; AVX512-FCP-NEXT: vpshufb %xmm7, %xmm3, %xmm3
-; AVX512-FCP-NEXT: vmovdqa64 %xmm7, %xmm24
-; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm18, %zmm3
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
+; AVX512-FCP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; AVX512-FCP-NEXT: vpshufb %xmm0, %xmm4, %xmm5
+; AVX512-FCP-NEXT: vpshufb %xmm7, %xmm4, %xmm4
+; AVX512-FCP-NEXT: vmovdqa64 %xmm7, %xmm26
+; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm29, %zmm4
; AVX512-FCP-NEXT: vpbroadcastd (%rax), %ymm5
; AVX512-FCP-NEXT: vpbroadcastd 4(%rax), %ymm6
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm6, %zmm5, %zmm30
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm30 = zmm30 ^ (zmm11 & (zmm30 ^ zmm3))
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm30 = zmm30 ^ (zmm1 & (zmm30 ^ zmm8))
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm6, %zmm5, %zmm29
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm29 = zmm29 ^ (zmm20 & (zmm29 ^ zmm4))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm29 = zmm29 ^ (zmm2 & (zmm29 ^ zmm8))
; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX512-FCP-NEXT: vmovdqa64 %ymm21, %ymm0
-; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm7, %ymm1
+; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm7, %ymm2
; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm21 # 32-byte Reload
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm21[2,2,2,2,6,6,6,6]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5,6,7,8],ymm1[9],ymm3[10,11],ymm1[12],ymm3[13,14,15]
-; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm7, %ymm3
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm21[2,2,2,2,6,6,6,6]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm2[1],ymm4[2,3],ymm2[4],ymm4[5,6,7,8],ymm2[9],ymm4[10,11],ymm2[12],ymm4[13,14,15]
+; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm7, %ymm4
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm21[1,1,1,1,5,5,5,5]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm5[2],ymm3[3,4],ymm5[5],ymm3[6,7,8,9],ymm5[10],ymm3[11,12],ymm5[13],ymm3[14,15]
-; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm19, %zmm3
-; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm9, %ymm1
-; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm18 # 32-byte Reload
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm18[2,2,2,2,6,6,6,6]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm5 = ymm1[0,1],ymm5[2],ymm1[3,4],ymm5[5],ymm1[6,7,8,9],ymm5[10],ymm1[11,12],ymm5[13],ymm1[14,15]
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm9[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm6 = ymm18[0,1,1,3,4,5,5,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm6[0,1],ymm1[2],ymm6[3,4],ymm1[5],ymm6[6,7,8,9],ymm1[10],ymm6[11,12],ymm1[13],ymm6[14,15]
-; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm13, %zmm1
-; AVX512-FCP-NEXT: vmovdqa 96(%rcx), %xmm11
-; AVX512-FCP-NEXT: vmovdqa 96(%rdx), %xmm13
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm13[0],xmm11[0],xmm13[1],xmm11[1],xmm13[2],xmm11[2],xmm13[3],xmm11[3]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7,8,9],ymm5[10],ymm4[11,12],ymm5[13],ymm4[14,15]
+; AVX512-FCP-NEXT: vpermt2q %zmm2, %zmm19, %zmm4
+; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm3, %ymm2
+; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm20 # 32-byte Reload
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm20[2,2,2,2,6,6,6,6]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm5 = ymm2[0,1],ymm5[2],ymm2[3,4],ymm5[5],ymm2[6,7,8,9],ymm5[10],ymm2[11,12],ymm5[13],ymm2[14,15]
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm6 = ymm20[0,1,1,3,4,5,5,7]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm6[0,1],ymm2[2],ymm6[3,4],ymm2[5],ymm6[6,7,8,9],ymm2[10],ymm6[11,12],ymm2[13],ymm6[14,15]
+; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm11, %zmm2
+; AVX512-FCP-NEXT: vmovdqa 96(%rcx), %xmm12
+; AVX512-FCP-NEXT: vmovdqa 96(%rdx), %xmm11
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm11[0],xmm12[0],xmm11[1],xmm12[1],xmm11[2],xmm12[2],xmm11[3],xmm12[3]
; AVX512-FCP-NEXT: vpshufb %xmm14, %xmm5, %xmm5
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm6 = [0,2,2,3,8,9,9,0]
-; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm6, %zmm12
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm9 = [0,2,2,3,8,9,9,0]
+; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm9, %zmm13
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm10 = [u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29]
-; AVX512-FCP-NEXT: vmovdqa64 %ymm27, %ymm2
-; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm2, %ymm5
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm8 = ymm28[3,3,3,3,7,7,7,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm8 = ymm8[0,1,2],ymm5[3],ymm8[4,5],ymm5[6],ymm8[7,8,9,10],ymm5[11],ymm8[12,13],ymm5[14],ymm8[15]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm3 ^ (mem & (zmm1 ^ zmm3))
+; AVX512-FCP-NEXT: vmovdqa64 %ymm22, %ymm5
+; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm5, %ymm5
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm6 = ymm28[3,3,3,3,7,7,7,7]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm8 = ymm6[0,1,2],ymm5[3],ymm6[4,5],ymm5[6],ymm6[7,8,9,10],ymm5[11],ymm6[12,13],ymm5[14],ymm6[15]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm4 ^ (zmm1 & (zmm2 ^ zmm4))
; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX512-FCP-NEXT: vprold $16, %ymm5, %ymm2
-; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm20 # 32-byte Reload
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm20[1,2,2,3,5,6,6,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7,8,9],ymm2[10],ymm3[11,12],ymm2[13],ymm3[14,15]
-; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm3 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
-; AVX512-FCP-NEXT: vmovdqa64 %ymm5, %ymm31
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm20[0,0,2,1,4,4,6,5]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3],ymm3[4,5],ymm5[6],ymm3[7,8,9,10],ymm5[11],ymm3[12,13],ymm5[14],ymm3[15]
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm0 = [2,2,3,3,10,9,11,10]
-; AVX512-FCP-NEXT: vpermt2q %zmm2, %zmm0, %zmm3
-; AVX512-FCP-NEXT: vmovdqa 96(%rsi), %xmm2
-; AVX512-FCP-NEXT: vmovdqa 96(%rdi), %xmm4
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; AVX512-FCP-NEXT: vpshufb %xmm15, %xmm5, %xmm5
+; AVX512-FCP-NEXT: vprold $16, %ymm5, %ymm1
+; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm17 # 32-byte Reload
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm17[1,2,2,3,5,6,6,7]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm4[0,1],ymm1[2],ymm4[3,4],ymm1[5],ymm4[6,7,8,9],ymm1[10],ymm4[11,12],ymm1[13],ymm4[14,15]
+; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm4 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
+; AVX512-FCP-NEXT: vmovdqa64 %ymm5, %ymm28
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm17[0,0,2,1,4,4,6,5]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3],ymm4[4,5],ymm5[6],ymm4[7,8,9,10],ymm5[11],ymm4[12,13],ymm5[14],ymm4[15]
+; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm27, %zmm4
+; AVX512-FCP-NEXT: vmovdqa 96(%rsi), %xmm1
+; AVX512-FCP-NEXT: vmovdqa 96(%rdi), %xmm5
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
+; AVX512-FCP-NEXT: vpshufb %xmm15, %xmm6, %xmm6
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm19 = [2,1,3,3,8,8,9,9]
-; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm19, %zmm8
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm17 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm12 ^ (zmm17 & (zmm8 ^ zmm12))
+; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm19, %zmm8
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm13 ^ (zmm23 & (zmm8 ^ zmm13))
; AVX512-FCP-NEXT: vbroadcasti32x4 {{.*#+}} ymm22 = [26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
; AVX512-FCP-NEXT: # ymm22 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT: vmovdqa64 %ymm22, %ymm0
-; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm5, %ymm0
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm25[3,3,3,3,7,7,7,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm12 = ymm0[0,1],ymm5[2],ymm0[3,4],ymm5[5],ymm0[6,7,8,9],ymm5[10],ymm0[11,12],ymm5[13],ymm0[14,15]
-; AVX512-FCP-NEXT: vpermd %zmm16, %zmm23, %zmm16
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm16 = zmm16 ^ (mem & (zmm16 ^ zmm3))
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm16 = zmm16 ^ (mem & (zmm16 ^ zmm1))
-; AVX512-FCP-NEXT: vmovdqa 32(%rcx), %xmm1
+; AVX512-FCP-NEXT: vmovdqa64 %ymm18, %ymm6
+; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm6, %ymm0
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm6 = ymm25[3,3,3,3,7,7,7,7]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm6 = ymm0[0,1],ymm6[2],ymm0[3,4],ymm6[5],ymm0[6,7,8,9],ymm6[10],ymm0[11,12],ymm6[13],ymm0[14,15]
+; AVX512-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,5,4,0,0,6,5,0,0,5,4,0,0,6,5,0]
+; AVX512-FCP-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512-FCP-NEXT: vpermd %zmm16, %zmm0, %zmm16
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm16 = zmm16 ^ (mem & (zmm16 ^ zmm4))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm16 = zmm16 ^ (mem & (zmm16 ^ zmm2))
+; AVX512-FCP-NEXT: vmovdqa 32(%rcx), %xmm2
+; AVX512-FCP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %xmm0
; AVX512-FCP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm1, %xmm28
-; AVX512-FCP-NEXT: vpshufb %xmm14, %xmm0, %xmm5
-; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm1 = ymm9[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm18[3,3,3,3,7,7,7,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5,6,7,8],ymm1[9],ymm3[10,11],ymm1[12],ymm3[13,14,15]
-; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm6, %zmm0
-; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm7, %ymm1
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; AVX512-FCP-NEXT: vpshufb %xmm14, %xmm0, %xmm4
+; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm2 = ymm3[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm20[3,3,3,3,7,7,7,7]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5,6,7,8],ymm2[9],ymm3[10,11],ymm2[12],ymm3[13,14,15]
+; AVX512-FCP-NEXT: vpermt2q %zmm4, %zmm9, %zmm0
+; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm7, %ymm2
; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm21[3,3,3,3,7,7,7,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm14 = ymm3[0,1,2],ymm1[3],ymm3[4,5],ymm1[6],ymm3[7,8,9,10],ymm1[11],ymm3[12,13],ymm1[14],ymm3[15]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm14 = ymm3[0,1,2],ymm2[3],ymm3[4,5],ymm2[6],ymm3[7,8,9,10],ymm2[11],ymm3[12,13],ymm2[14],ymm3[15]
; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %xmm10
; AVX512-FCP-NEXT: vmovdqa 32(%rsi), %xmm9
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; AVX512-FCP-NEXT: vpshufb %xmm15, %xmm1, %xmm1
-; AVX512-FCP-NEXT: vmovdqa64 %xmm15, %xmm25
-; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm19, %zmm14
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
+; AVX512-FCP-NEXT: vpshufb %xmm15, %xmm2, %xmm2
+; AVX512-FCP-NEXT: vmovdqa64 %xmm15, %xmm27
+; AVX512-FCP-NEXT: vpermt2q %zmm2, %zmm19, %zmm14
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm21 = [2,2,2,3,8,8,8,9]
-; AVX512-FCP-NEXT: vmovdqa 96(%r9), %xmm6
-; AVX512-FCP-NEXT: vmovdqa 96(%r8), %xmm5
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm24, %xmm18
-; AVX512-FCP-NEXT: vmovdqa64 %xmm24, %xmm1
-; AVX512-FCP-NEXT: vpshufb %xmm1, %xmm7, %xmm15
-; AVX512-FCP-NEXT: vpermt2q %zmm15, %zmm21, %zmm12
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm14 = zmm0 ^ (zmm17 & (zmm14 ^ zmm0))
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm15 = [6,7,3,3,7,7,6,7]
-; AVX512-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: vmovdqa 96(%r9), %xmm4
+; AVX512-FCP-NEXT: vmovdqa 96(%r8), %xmm3
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; AVX512-FCP-NEXT: vmovdqa64 %xmm26, %xmm13
+; AVX512-FCP-NEXT: vpshufb %xmm13, %xmm7, %xmm15
+; AVX512-FCP-NEXT: vpermt2q %zmm15, %zmm21, %zmm6
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm14 = zmm0 ^ (zmm23 & (zmm14 ^ zmm0))
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm18 = [6,7,3,3,7,7,6,7]
+; AVX512-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm18, %ymm0 # 32-byte Folded Reload
; AVX512-FCP-NEXT: vpbroadcastd 96(%rax), %ymm19
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm19, %zmm0, %zmm19
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm24 = [65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm19 = zmm19 ^ (zmm24 & (zmm19 ^ zmm12))
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm23 = [0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm19 = zmm19 ^ (zmm23 & (zmm19 ^ zmm8))
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
-; AVX512-FCP-NEXT: vprold $16, %xmm2, %xmm0
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm4[1,1,2,3]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm23 = [65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm19 = zmm19 ^ (zmm23 & (zmm19 ^ zmm6))
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm26 = [0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm19 = zmm19 ^ (zmm26 & (zmm19 ^ zmm8))
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
+; AVX512-FCP-NEXT: vprold $16, %xmm1, %xmm0
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm1 = xmm5[1,1,2,3]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3,4],xmm0[5],xmm1[6,7]
; AVX512-FCP-NEXT: vmovdqa64 %ymm22, %ymm1
-; AVX512-FCP-NEXT: vmovdqa64 %ymm31, %ymm2
-; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm2, %ymm1
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm2 = ymm20[3,3,3,3,7,7,7,7]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
+; AVX512-FCP-NEXT: vmovdqa64 %ymm28, %ymm5
+; AVX512-FCP-NEXT: vpshufb %ymm1, %ymm5, %ymm1
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm17[3,3,3,3,7,7,7,7]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm5[2],ymm1[3,4],ymm5[5],ymm1[6,7,8,9],ymm5[10],ymm1[11,12],ymm5[13],ymm1[14,15]
; AVX512-FCP-NEXT: vmovdqa 32(%r9), %xmm8
-; AVX512-FCP-NEXT: vmovdqa 32(%r8), %xmm4
-; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm18, %xmm12
-; AVX512-FCP-NEXT: vpshufb %xmm12, %xmm2, %xmm12
-; AVX512-FCP-NEXT: vpermt2q %zmm12, %zmm21, %zmm1
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm12 = [4,5,2,3,4,5,6,7,8,9,10,11,10,11,8,9]
-; AVX512-FCP-NEXT: vpshufb %xmm12, %xmm3, %xmm3
-; AVX512-FCP-NEXT: vmovdqa64 %xmm12, %xmm31
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm17 = [0,0,0,1,8,8,9,0]
-; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm17, %zmm0
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm13[4],xmm11[4],xmm13[5],xmm11[5],xmm13[6],xmm11[6],xmm13[7],xmm11[7]
-; AVX512-FCP-NEXT: vpbroadcastq {{.*#+}} xmm22 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm22, %xmm12
-; AVX512-FCP-NEXT: vpshufb %xmm12, %xmm11, %xmm11
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[1,1,2,2]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm11 = xmm13[0],xmm11[1],xmm13[2,3],xmm11[4],xmm13[5,6],xmm11[7]
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,2,3,3,4,5,6,7]
-; AVX512-FCP-NEXT: vmovdqa64 %zmm26, %zmm21
-; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm26, %zmm11
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm26 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm11 = zmm0 ^ (zmm26 & (zmm11 ^ zmm0))
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
-; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm20 # 32-byte Reload
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} ymm20 = ymm20 ^ (ymm24 & (ymm20 ^ ymm0))
-; AVX512-FCP-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} xmm18 = [0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm18, %xmm5
-; AVX512-FCP-NEXT: vpshufb %xmm5, %xmm7, %xmm5
-; AVX512-FCP-NEXT: vmovdqa64 %xmm25, %xmm12
-; AVX512-FCP-NEXT: vpshufb %xmm12, %xmm3, %xmm3
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm25 = [8,9,9,0,0,0,1,1]
-; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm25, %zmm3
-; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX512-FCP-NEXT: vpshufhw {{.*#+}} ymm5 = ymm6[0,1,2,3,5,5,7,6,8,9,10,11,13,13,15,14]
-; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm5 = ymm5[3,3,3,3]
-; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,24,25,22,23,22,23,u,u,u,u,u,u,u,u]
-; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm6 = ymm6[2,2,2,2]
+; AVX512-FCP-NEXT: vmovdqa 32(%r8), %xmm6
+; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm8[0],xmm6[1],xmm8[1],xmm6[2],xmm8[2],xmm6[3],xmm8[3]
+; AVX512-FCP-NEXT: vpshufb %xmm13, %xmm5, %xmm15
+; AVX512-FCP-NEXT: vpermt2q %zmm15, %zmm21, %zmm1
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm13 = [4,5,2,3,4,5,6,7,8,9,10,11,10,11,8,9]
+; AVX512-FCP-NEXT: vpshufb %xmm13, %xmm2, %xmm2
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm21 = [0,0,0,1,8,8,9,0]
+; AVX512-FCP-NEXT: vpermt2q %zmm2, %zmm21, %zmm0
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm11[4],xmm12[4],xmm11[5],xmm12[5],xmm11[6],xmm12[6],xmm11[7],xmm12[7]
+; AVX512-FCP-NEXT: vpbroadcastq {{.*#+}} xmm20 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
+; AVX512-FCP-NEXT: vmovdqa64 %xmm20, %xmm15
+; AVX512-FCP-NEXT: vpshufb %xmm15, %xmm12, %xmm12
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[1,1,2,2]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm12 = xmm11[0],xmm12[1],xmm11[2,3],xmm12[4],xmm11[5,6],xmm12[7]
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,2,3,3,4,5,6,7]
+; AVX512-FCP-NEXT: vmovdqa64 %zmm24, %zmm25
+; AVX512-FCP-NEXT: vpermt2q %zmm2, %zmm24, %zmm12
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm28 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm12 = zmm0 ^ (zmm28 & (zmm12 ^ zmm0))
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm22 # 32-byte Reload
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} ymm22 = ymm22 ^ (ymm23 & (ymm22 ^ ymm2))
+; AVX512-FCP-NEXT: vextracti64x4 $1, %zmm2, %ymm2
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} xmm24 = [0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
+; AVX512-FCP-NEXT: vmovdqa64 %xmm24, %xmm3
+; AVX512-FCP-NEXT: vpshufb %xmm3, %xmm7, %xmm3
+; AVX512-FCP-NEXT: vmovdqa64 %xmm27, %xmm15
+; AVX512-FCP-NEXT: vpshufb %xmm15, %xmm0, %xmm0
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm27 = [8,9,9,0,0,0,1,1]
+; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm27, %zmm0
+; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX512-FCP-NEXT: vpshufhw {{.*#+}} ymm3 = ymm4[0,1,2,3,5,5,7,6,8,9,10,11,13,13,15,14]
+; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm3 = ymm3[3,3,3,3]
+; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,24,25,22,23,22,23,u,u,u,u,u,u,u,u]
+; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,2,2,2]
; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX512-FCP-NEXT: vpshufb %xmm12, %xmm7, %xmm7
-; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm13 = ymm7[0,0,1,1]
+; AVX512-FCP-NEXT: vpshufb %xmm15, %xmm7, %xmm7
+; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm11 = ymm7[0,0,1,1]
; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX512-FCP-NEXT: vpshufb %xmm12, %xmm7, %xmm7
-; AVX512-FCP-NEXT: vmovdqa64 %xmm12, %xmm27
-; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm12 = ymm7[0,0,1,1]
-; AVX512-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm7 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: vpshufb %xmm15, %xmm7, %xmm7
+; AVX512-FCP-NEXT: vmovdqa64 %xmm15, %xmm31
+; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm17 = ymm7[0,0,1,1]
+; AVX512-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm18, %ymm7 # 32-byte Folded Reload
; AVX512-FCP-NEXT: vpbroadcastd 32(%rax), %ymm15
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm15, %zmm7, %zmm7
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (zmm24 & (zmm7 ^ zmm1))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (zmm23 & (zmm7 ^ zmm1))
; AVX512-FCP-NEXT: vpbroadcastd 100(%rax), %ymm1
; AVX512-FCP-NEXT: vpbroadcastd 104(%rax), %ymm15
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm15, %zmm1, %zmm15
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm24 = [65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm15 = zmm15 ^ (zmm24 & (zmm15 ^ zmm3))
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (zmm23 & (zmm7 ^ zmm14))
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm15 = zmm15 ^ (zmm1 & (zmm15 ^ zmm11))
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm18 = [65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm15 = zmm15 ^ (zmm18 & (zmm15 ^ zmm0))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (zmm26 & (zmm7 ^ zmm14))
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm15 = zmm15 ^ (zmm0 & (zmm15 ^ zmm12))
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
; AVX512-FCP-NEXT: vprold $16, %xmm9, %xmm9
; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[1,1,2,3]
; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0,1],xmm9[2],xmm10[3,4],xmm9[5],xmm10[6,7]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm31, %xmm14
-; AVX512-FCP-NEXT: vpshufb %xmm14, %xmm3, %xmm3
-; AVX512-FCP-NEXT: vpermt2q %zmm3, %zmm17, %zmm9
-; AVX512-FCP-NEXT: vmovdqa64 %xmm22, %xmm3
-; AVX512-FCP-NEXT: vmovdqa64 %xmm28, %xmm10
-; AVX512-FCP-NEXT: vpshufb %xmm3, %xmm10, %xmm3
-; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm10 = xmm11[4],xmm10[4],xmm11[5],xmm10[5],xmm11[6],xmm10[6],xmm11[7],xmm10[7]
-; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[1,1,2,2]
-; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm3 = xmm11[0],xmm3[1],xmm11[2,3],xmm3[4],xmm11[5,6],xmm3[7]
+; AVX512-FCP-NEXT: vpshufb %xmm13, %xmm1, %xmm1
+; AVX512-FCP-NEXT: vpermt2q %zmm1, %zmm21, %zmm9
+; AVX512-FCP-NEXT: vmovdqa64 %xmm20, %xmm1
+; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX512-FCP-NEXT: vpshufb %xmm1, %xmm10, %xmm1
+; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm10 = xmm12[4],xmm10[4],xmm12[5],xmm10[5],xmm12[6],xmm10[6],xmm12[7],xmm10[7]
+; AVX512-FCP-NEXT: vpshufd {{.*#+}} xmm12 = xmm12[1,1,2,2]
+; AVX512-FCP-NEXT: vpblendw {{.*#+}} xmm1 = xmm12[0],xmm1[1],xmm12[2,3],xmm1[4],xmm12[5,6],xmm1[7]
; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,2,3,3,4,5,6,7]
-; AVX512-FCP-NEXT: vpermt2q %zmm10, %zmm21, %zmm3
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm3 = zmm9 ^ (zmm26 & (zmm3 ^ zmm9))
-; AVX512-FCP-NEXT: vmovdqa64 %xmm18, %xmm9
-; AVX512-FCP-NEXT: vpshufb %xmm9, %xmm2, %xmm2
-; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm27, %xmm8
-; AVX512-FCP-NEXT: vpshufb %xmm8, %xmm4, %xmm4
-; AVX512-FCP-NEXT: vpermt2q %zmm2, %zmm25, %zmm4
-; AVX512-FCP-NEXT: vpbroadcastd 36(%rax), %ymm2
+; AVX512-FCP-NEXT: vpermt2q %zmm10, %zmm25, %zmm1
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm9 ^ (zmm28 & (zmm1 ^ zmm9))
+; AVX512-FCP-NEXT: vmovdqa64 %xmm24, %xmm9
+; AVX512-FCP-NEXT: vpshufb %xmm9, %xmm5, %xmm5
+; AVX512-FCP-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm6[4],xmm8[4],xmm6[5],xmm8[5],xmm6[6],xmm8[6],xmm6[7],xmm8[7]
+; AVX512-FCP-NEXT: vmovdqa64 %xmm31, %xmm8
+; AVX512-FCP-NEXT: vpshufb %xmm8, %xmm6, %xmm6
+; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm27, %zmm6
+; AVX512-FCP-NEXT: vpbroadcastd 36(%rax), %ymm5
; AVX512-FCP-NEXT: vpbroadcastd 40(%rax), %ymm8
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm2, %zmm2
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm24 & (zmm2 ^ zmm4))
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm1 & (zmm2 ^ zmm3))
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
-; AVX512-FCP-NEXT: vpternlogd {{.*#+}} ymm0 = mem ^ (ymm1 & (ymm0 ^ mem))
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} ymm5 = ymm5 ^ (ymm1 & (ymm5 ^ ymm0))
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm4 = [65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} ymm6 = ymm6 ^ (ymm4 & (ymm6 ^ ymm20))
-; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm6[0,1,2,3],zmm5[0,1,2,3]
-; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
-; AVX512-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm3 # 64-byte Folded Reload
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm3 = zmm3 ^ (mem & (zmm3 ^ zmm0))
-; AVX512-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm13, %zmm0 # 32-byte Folded Reload
-; AVX512-FCP-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm0 # 64-byte Folded Reload
-; AVX512-FCP-NEXT: # zmm0 = (zmm0 & zmm4) | mem
-; AVX512-FCP-NEXT: vmovdqa64 %zmm4, %zmm5
-; AVX512-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm12, %zmm4 # 32-byte Folded Reload
-; AVX512-FCP-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm5, %zmm4 # 64-byte Folded Reload
-; AVX512-FCP-NEXT: # zmm4 = (zmm4 & zmm5) | mem
-; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX512-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX512-FCP-NEXT: # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm5, %zmm5
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (zmm18 & (zmm5 ^ zmm6))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (zmm0 & (zmm5 ^ zmm1))
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
+; AVX512-FCP-NEXT: vpternlogd {{.*#+}} ymm2 = mem ^ (ymm0 & (ymm2 ^ mem))
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} ymm3 = ymm3 ^ (ymm0 & (ymm3 ^ ymm2))
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm6 = [65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} ymm4 = ymm4 ^ (ymm6 & (ymm4 ^ ymm22))
+; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm1 = zmm4[0,1,2,3],zmm3[0,1,2,3]
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm2 = [0,13,0,0,0,14,0,0,14,0,0,0,15,0,0,15]
+; AVX512-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm2 # 64-byte Folded Reload
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm1))
+; AVX512-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm11, %zmm1 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm6, %zmm1 # 64-byte Folded Reload
+; AVX512-FCP-NEXT: # zmm1 = (zmm1 & zmm6) | mem
+; AVX512-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm17, %zmm3 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm6, %zmm3 # 64-byte Folded Reload
+; AVX512-FCP-NEXT: # zmm3 = (zmm3 & zmm6) | mem
+; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX512-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX512-FCP-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX512-FCP-NEXT: vpunpckhwd (%rsp), %xmm6, %xmm6 # 16-byte Folded Reload
; AVX512-FCP-NEXT: # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
@@ -13789,9 +13779,8 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX512-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
; AVX512-FCP-NEXT: # xmm9 = xmm9[4],mem[4],xmm9[5],mem[5],xmm9[6],mem[6],xmm9[7],mem[7]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm31, %xmm10
-; AVX512-FCP-NEXT: vpshufb %xmm10, %xmm6, %xmm6
-; AVX512-FCP-NEXT: vpshufb %xmm10, %xmm9, %xmm9
+; AVX512-FCP-NEXT: vpshufb %xmm13, %xmm6, %xmm6
+; AVX512-FCP-NEXT: vpshufb %xmm13, %xmm9, %xmm9
; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX512-FCP-NEXT: vpunpcklwd {{.*#+}} ymm10 = ymm10[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
; AVX512-FCP-NEXT: vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
@@ -13809,8 +13798,8 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm14 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535]
; AVX512-FCP-NEXT: vpandn %ymm13, %ymm14, %ymm13
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm13, %zmm12, %zmm12
-; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,2,3,3,4,5,6,7]
-; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,0,2,1]
+; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,2,3,3,4,5,6,7]
+; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,0,2,1]
; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,0,1,3]
; AVX512-FCP-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,2,3,3,4,5,6,7]
; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm8 = ymm8[0,0,2,1]
@@ -13820,43 +13809,43 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm14 = zmm14 ^ (mem & (zmm14 ^ zmm13))
-; AVX512-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm5, %zmm5 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm4 # 32-byte Folded Reload
; AVX512-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm6, %zmm6 # 32-byte Folded Reload
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm5 ^ (zmm26 & (zmm6 ^ zmm5))
-; AVX512-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm5 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm4 ^ (zmm28 & (zmm6 ^ zmm4))
+; AVX512-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm4 # 32-byte Folded Reload
; AVX512-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm8 # 32-byte Folded Reload
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm5 ^ (zmm26 & (zmm8 ^ zmm5))
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm5 = [0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm0 ^ (zmm5 & (zmm6 ^ zmm0))
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm4 ^ (zmm5 & (zmm8 ^ zmm4))
-; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm10, %zmm0, %zmm0
-; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm11, %zmm4, %zmm4
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm0 ^ (mem & (zmm4 ^ zmm0))
-; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512-FCP-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
-; AVX512-FCP-NEXT: # zmm0 = zmm0 | (zmm1 & mem)
-; AVX512-FCP-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm12 # 64-byte Folded Reload
-; AVX512-FCP-NEXT: # zmm12 = zmm12 | (zmm1 & mem)
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm12 = zmm12 ^ (mem & (zmm12 ^ zmm4))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm4 ^ (zmm28 & (zmm8 ^ zmm4))
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm4 = [0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm1 ^ (zmm4 & (zmm6 ^ zmm1))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm3 ^ (zmm4 & (zmm8 ^ zmm3))
+; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm10, %zmm1, %zmm1
+; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm11, %zmm3, %zmm3
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm3 = zmm1 ^ (mem & (zmm3 ^ zmm1))
+; AVX512-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512-FCP-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 # 64-byte Folded Reload
+; AVX512-FCP-NEXT: # zmm1 = zmm1 | (zmm0 & mem)
+; AVX512-FCP-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm12 # 64-byte Folded Reload
+; AVX512-FCP-NEXT: # zmm12 = zmm12 | (zmm0 & mem)
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm12 = zmm12 ^ (mem & (zmm12 ^ zmm3))
; AVX512-FCP-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512-FCP-NEXT: vmovdqa64 %zmm12, 320(%rax)
-; AVX512-FCP-NEXT: vmovdqa64 %zmm2, 256(%rax)
+; AVX512-FCP-NEXT: vmovdqa64 %zmm5, 256(%rax)
; AVX512-FCP-NEXT: vmovdqa64 %zmm7, 192(%rax)
; AVX512-FCP-NEXT: vmovdqa64 %zmm16, 128(%rax)
-; AVX512-FCP-NEXT: vmovdqa64 %zmm30, (%rax)
-; AVX512-FCP-NEXT: vmovdqa64 %zmm29, 448(%rax)
+; AVX512-FCP-NEXT: vmovdqa64 %zmm29, (%rax)
+; AVX512-FCP-NEXT: vmovdqa64 %zmm30, 448(%rax)
; AVX512-FCP-NEXT: vmovdqa64 %zmm15, 704(%rax)
; AVX512-FCP-NEXT: vmovdqa64 %zmm19, 640(%rax)
-; AVX512-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
-; AVX512-FCP-NEXT: vmovaps %zmm1, 576(%rax)
+; AVX512-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512-FCP-NEXT: vmovaps %zmm0, 576(%rax)
; AVX512-FCP-NEXT: vmovdqa64 %zmm14, 384(%rax)
; AVX512-FCP-NEXT: vmovdqa64 %zmm8, 64(%rax)
; AVX512-FCP-NEXT: vmovdqa64 %zmm6, 512(%rax)
-; AVX512-FCP-NEXT: vmovdqa64 %zmm3, 832(%rax)
-; AVX512-FCP-NEXT: vmovdqa64 %zmm0, 768(%rax)
-; AVX512-FCP-NEXT: addq $1544, %rsp # imm = 0x608
+; AVX512-FCP-NEXT: vmovdqa64 %zmm2, 832(%rax)
+; AVX512-FCP-NEXT: vmovdqa64 %zmm1, 768(%rax)
+; AVX512-FCP-NEXT: addq $1480, %rsp # imm = 0x5C8
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
@@ -14563,7 +14552,7 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX512DQ-FCP-LABEL: store_i16_stride7_vf64:
; AVX512DQ-FCP: # %bb.0:
-; AVX512DQ-FCP-NEXT: subq $1544, %rsp # imm = 0x608
+; AVX512DQ-FCP-NEXT: subq $1480, %rsp # imm = 0x5C8
; AVX512DQ-FCP-NEXT: vmovdqa 96(%rcx), %ymm3
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128]
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm3, %ymm0
@@ -14590,48 +14579,48 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa 64(%rcx), %ymm0
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm0, %ymm14
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm0, %ymm29
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm0, %ymm27
; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdx), %ymm0
; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm0, %ymm15
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm0, %ymm18
-; AVX512DQ-FCP-NEXT: vpor %ymm14, %ymm15, %ymm14
-; AVX512DQ-FCP-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm0, %ymm29
+; AVX512DQ-FCP-NEXT: vpor %ymm14, %ymm15, %ymm13
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa 64(%rsi), %ymm0
; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm0, %ymm14
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm0, %ymm27
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm0, %ymm22
; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdi), %ymm0
; AVX512DQ-FCP-NEXT: vpshufb %ymm11, %ymm0, %ymm15
; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm0, %ymm28
-; AVX512DQ-FCP-NEXT: vpor %ymm14, %ymm15, %ymm14
-; AVX512DQ-FCP-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpor %ymm14, %ymm15, %ymm13
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa (%r9), %ymm0
; AVX512DQ-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm0, %ymm14
; AVX512DQ-FCP-NEXT: vmovdqa (%r8), %ymm0
; AVX512DQ-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vpshufb %ymm12, %ymm0, %ymm15
-; AVX512DQ-FCP-NEXT: vpor %ymm14, %ymm15, %ymm14
-; AVX512DQ-FCP-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpor %ymm14, %ymm15, %ymm13
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %ymm0
; AVX512DQ-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm0, %ymm14
; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %ymm0
; AVX512DQ-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm0, %ymm15
-; AVX512DQ-FCP-NEXT: vpor %ymm14, %ymm15, %ymm14
-; AVX512DQ-FCP-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpor %ymm14, %ymm15, %ymm13
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %ymm0
; AVX512DQ-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm0, %ymm14
+; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm0, %ymm15
; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %ymm0
; AVX512DQ-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %ymm11, %ymm0, %ymm15
-; AVX512DQ-FCP-NEXT: vpor %ymm14, %ymm15, %ymm14
-; AVX512DQ-FCP-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpshufb %ymm11, %ymm0, %ymm13
+; AVX512DQ-FCP-NEXT: vpor %ymm15, %ymm13, %ymm13
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rcx), %ymm15
; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm15, %ymm6
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %ymm13
-; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm13, %ymm8
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %ymm14
+; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm14, %ymm8
; AVX512DQ-FCP-NEXT: vpor %ymm6, %ymm8, %ymm6
; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rsi), %ymm0
@@ -14649,32 +14638,32 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm12 = ymm1[0,1,1,3,4,5,5,7]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm11 = ymm12[0,1],ymm11[2],ymm12[3,4],ymm11[5],ymm12[6,7,8,9],ymm11[10],ymm12[11,12],ymm11[13],ymm12[14,15]
; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm11 = ymm11[2,1,3,2]
-; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm14 = [18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
-; AVX512DQ-FCP-NEXT: # ymm14 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm14, %ymm2, %ymm12
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm14, %ymm17
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm14 = ymm7[1,1,1,1,5,5,5,5]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm12 = ymm12[0,1],ymm14[2],ymm12[3,4],ymm14[5],ymm12[6,7,8,9],ymm14[10],ymm12[11,12],ymm14[13],ymm12[14,15]
+; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm13 = [18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
+; AVX512DQ-FCP-NEXT: # ymm13 = mem[0,1,0,1]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm13, %ymm2, %ymm12
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm13, %ymm17
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm13 = ymm7[1,1,1,1,5,5,5,5]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm12 = ymm12[0,1],ymm13[2],ymm12[3,4],ymm13[5],ymm12[6,7,8,9],ymm13[10],ymm12[11,12],ymm13[13],ymm12[14,15]
; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm12 = ymm12[2,2,2,3]
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm11, %zmm16, %zmm11
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm12, %zmm5, %zmm12
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm12 = zmm11 ^ (mem & (zmm12 ^ zmm11))
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm14 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535]
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm13 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535]
; AVX512DQ-FCP-NEXT: vmovdqa 96(%r8), %ymm5
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm11 = ymm5[u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm5[14,15,u,u,u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm5[16,17,u,u,u,u],zero,zero
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} ymm11 = ymm11 | (ymm12 & ymm14)
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} ymm11 = ymm11 | (ymm12 & ymm13)
; AVX512DQ-FCP-NEXT: vmovdqa 96(%r9), %ymm6
-; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm6, %ymm4
; AVX512DQ-FCP-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm6, %ymm4
; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm16 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} ymm4 = ymm4 | (ymm11 & ymm16)
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm16, %zmm30
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm16, %zmm20
; AVX512DQ-FCP-NEXT: vextracti64x4 $1, %zmm12, %ymm11
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm12 = [0,4,0,0,0,5,0,0]
; AVX512DQ-FCP-NEXT: vpermd %ymm5, %ymm12, %ymm12
; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm16 = [65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0]
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} ymm12 = ymm12 ^ (ymm16 & (ymm12 ^ ymm11))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm16, %zmm20
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm16, %zmm18
; AVX512DQ-FCP-NEXT: vprold $16, %ymm6, %ymm11
; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm11 = ymm11[2,2,2,2]
; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm16 = [65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535]
@@ -14685,19 +14674,19 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm11 = [0,1,4,5,4,5,5,7]
; AVX512DQ-FCP-NEXT: vmovdqa 96(%rax), %ymm4
; AVX512DQ-FCP-NEXT: vpermd %ymm4, %ymm11, %ymm11
-; AVX512DQ-FCP-NEXT: vpandn %ymm11, %ymm14, %ymm11
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm14, %zmm22
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm12 = [12,13,128,128,128,128,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm12, %ymm4, %ymm4
+; AVX512DQ-FCP-NEXT: vpandn %ymm11, %ymm13, %ymm11
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm13, %zmm12
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm13 = [12,13,128,128,128,128,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm13, %ymm4, %ymm4
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm11, %zmm4, %zmm4
; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm14 = [u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm14, %ymm2, %ymm4
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm6 = [u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm2, %ymm4
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm11 = ymm7[3,3,3,3,7,7,7,7]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm11[0,1,2],ymm4[3],ymm11[4,5],ymm4[6],ymm11[7,8,9,10],ymm4[11],ymm11[12,13],ymm4[14],ymm11[15]
-; AVX512DQ-FCP-NEXT: vpmovsxdq {{.*#+}} ymm6 = [151522058,0,421010202,421010202]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm2, %ymm2
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm6, %ymm23
+; AVX512DQ-FCP-NEXT: vpmovsxdq {{.*#+}} ymm11 = [151522058,0,421010202,421010202]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm11, %ymm2, %ymm2
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm11, %ymm24
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm7 = ymm7[2,2,2,2,6,6,6,6]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm7[0],ymm2[1],ymm7[2,3],ymm2[4],ymm7[5,6,7,8],ymm2[9],ymm7[10,11],ymm2[12],ymm7[13,14,15]
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm7 = [0,2,2,3,10,9,11,11]
@@ -14707,16 +14696,15 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm11, %ymm16
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm11 = ymm1[3,3,3,3,7,7,7,7]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm11[0],ymm4[1],ymm11[2,3],ymm4[4],ymm11[5,6,7,8],ymm4[9],ymm11[10,11],ymm4[12],ymm11[13,14,15]
-; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm6 = [22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27,22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27]
-; AVX512DQ-FCP-NEXT: # ymm6 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm6, %ymm24
+; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm11 = [22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27,22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27]
+; AVX512DQ-FCP-NEXT: # ymm11 = mem[0,1,0,1]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm11, %ymm3, %ymm3
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,2,2,2,6,6,6,6]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0,1],ymm1[2],ymm3[3,4],ymm1[5],ymm3[6,7,8,9],ymm1[10],ymm3[11,12],ymm1[13],ymm3[14,15]
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm19 = [2,2,2,3,8,10,10,11]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm4, %zmm19, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm11 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm2 ^ (zmm11 & (zmm1 ^ zmm2))
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm23 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm2 ^ (zmm23 & (zmm1 ^ zmm2))
; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm1 = [5,0,0,0,6,0,0,6]
; AVX512DQ-FCP-NEXT: vpermd %ymm5, %ymm1, %ymm1
@@ -14726,45 +14714,45 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vpandn %ymm1, %ymm3, %ymm1
; AVX512DQ-FCP-NEXT: vmovdqa 64(%rax), %ymm2
; AVX512DQ-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %ymm12, %ymm2, %ymm2
+; AVX512DQ-FCP-NEXT: vpshufb %ymm13, %ymm2, %ymm2
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT: vpbroadcastd 8(%rax), %ymm1
; AVX512DQ-FCP-NEXT: vpandn %ymm1, %ymm3, %ymm1
; AVX512DQ-FCP-NEXT: vmovdqa (%rax), %ymm2
; AVX512DQ-FCP-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %ymm12, %ymm2, %ymm2
+; AVX512DQ-FCP-NEXT: vpshufb %ymm13, %ymm2, %ymm2
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vpbroadcastd {{.*#+}} ymm6 = [18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm10, %ymm1
+; AVX512DQ-FCP-NEXT: vpbroadcastd {{.*#+}} ymm5 = [18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm10, %ymm1
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm2 = ymm8[0,0,2,1,4,4,6,5]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7,8,9,10],ymm2[11],ymm1[12,13],ymm2[14],ymm1[15]
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm2 = [0,1,0,3,10,10,11,11]
; AVX512DQ-FCP-NEXT: vpermi2q %zmm1, %zmm21, %zmm2
; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %ymm14, %ymm0, %ymm1
+; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm0, %ymm1
; AVX512DQ-FCP-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5],ymm1[6],ymm2[7,8,9,10],ymm1[11],ymm2[12,13],ymm1[14],ymm2[15]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm23, %ymm5
-; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm0, %ymm2
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm24, %ymm13
+; AVX512DQ-FCP-NEXT: vpshufb %ymm13, %ymm0, %ymm2
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm9[2,2,2,2,6,6,6,6]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5,6,7,8],ymm2[9],ymm3[10,11],ymm2[12],ymm3[13,14,15]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm7, %zmm2
; AVX512DQ-FCP-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm16, %ymm7
-; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm15, %ymm1
-; AVX512DQ-FCP-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm13[3,3,3,3,7,7,7,7]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm16, %ymm6
+; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm15, %ymm1
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm14[3,3,3,3,7,7,7,7]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5,6,7,8],ymm1[9],ymm3[10,11],ymm1[12],ymm3[13,14,15]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm24, %ymm12
-; AVX512DQ-FCP-NEXT: vpshufb %ymm12, %ymm15, %ymm3
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm13[2,2,2,2,6,6,6,6]
+; AVX512DQ-FCP-NEXT: vmovdqa %ymm11, %ymm7
+; AVX512DQ-FCP-NEXT: vpshufb %ymm11, %ymm15, %ymm3
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm14[2,2,2,2,6,6,6,6]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7,8,9],ymm4[10],ymm3[11,12],ymm4[13],ymm3[14,15]
; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm19, %zmm0
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm2 ^ (zmm11 & (zmm0 ^ zmm2))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm2 ^ (zmm23 & (zmm0 ^ zmm2))
; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm1 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm2 = ymm8[3,3,3,3,7,7,7,7]
@@ -14772,348 +14760,343 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vprold $16, %ymm10, %ymm2
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm8[1,2,2,3,5,6,6,7]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7,8,9],ymm2[10],ymm3[11,12],ymm2[13],ymm3[14,15]
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm13 = [2,1,3,2,10,10,10,11]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm13, %zmm2
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm11 = [2,1,3,2,10,10,10,11]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm11, %zmm2
; AVX512DQ-FCP-NEXT: vmovdqa64 (%rax), %zmm16
-; AVX512DQ-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [14,13,8,8,15,14,8,15,14,13,8,8,15,14,8,15]
-; AVX512DQ-FCP-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3]
-; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm0 = [0,13,0,0,0,14,0,0,14,0,0,0,15,0,0,15]
; AVX512DQ-FCP-NEXT: vpermd %zmm16, %zmm0, %zmm0
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (zmm20 & (zmm0 ^ zmm2))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (zmm18 & (zmm0 ^ zmm2))
; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm27, %ymm0
-; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm23, %ymm21
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm22, %ymm15
+; AVX512DQ-FCP-NEXT: vpshufb %ymm13, %ymm15, %ymm1
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm24, %ymm21
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm2 = ymm28[2,2,2,2,6,6,6,6]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6,7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13,14,15]
; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm17, %ymm9
-; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm0, %ymm2
+; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm15, %ymm2
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm28[1,1,1,1,5,5,5,5]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm3 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7,8,9],ymm3[10],ymm2[11,12],ymm3[13],ymm2[14,15]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm19, %zmm3
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm29, %ymm8
-; AVX512DQ-FCP-NEXT: vpshufb %ymm12, %ymm8, %ymm1
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm24, %ymm10
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm2 = ymm18[2,2,2,2,6,6,6,6]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm8[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm18[0,1,1,3,4,5,5,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0,1],ymm2[2],ymm4[3,4],ymm2[5],ymm4[6,7,8,9],ymm2[10],ymm4[11,12],ymm2[13],ymm4[14,15]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm13, %zmm5
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm3 ^ (mem & (zmm5 ^ zmm3))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm25, %ymm0
-; AVX512DQ-FCP-NEXT: vmovdqu64 %ymm25, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FCP-NEXT: vprold $16, %ymm25, %ymm1
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7,8,9],ymm3[10],ymm2[11,12],ymm3[13],ymm2[14,15]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm19, %zmm2
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm27, %ymm0
+; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm0, %ymm1
+; AVX512DQ-FCP-NEXT: vmovdqa %ymm7, %ymm10
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm29[2,2,2,2,6,6,6,6]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm0[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm27, %ymm7
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm29[0,1,1,3,4,5,5,7]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7,8,9],ymm3[10],ymm4[11,12],ymm3[13],ymm4[14,15]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm11, %zmm4
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm2 ^ (zmm1 & (zmm4 ^ zmm2))
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm25, %ymm13
+; AVX512DQ-FCP-NEXT: vprold $16, %ymm25, %ymm2
; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm26, %ymm25
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm26[1,2,2,3,5,6,6,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0,1],ymm1[2],ymm3[3,4],ymm1[5],ymm3[6,7,8,9],ymm1[10],ymm3[11,12],ymm1[13],ymm3[14,15]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm0, %ymm3
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm26[0,0,2,1,4,4,6,5]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3],ymm3[4,5],ymm4[6],ymm3[7,8,9,10],ymm4[11],ymm3[12,13],ymm4[14],ymm3[15]
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm0 = [2,2,3,3,10,9,11,10]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm0, %zmm3
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7,8,9],ymm2[10],ymm3[11,12],ymm2[13],ymm3[14,15]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm5, %ymm13, %ymm3
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm13, %ymm18
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm26[0,0,2,1,4,4,6,5]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3],ymm3[4,5],ymm5[6],ymm3[7,8,9,10],ymm5[11],ymm3[12,13],ymm5[14],ymm3[15]
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm27 = [2,2,3,3,10,9,11,10]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm2, %zmm27, %zmm3
; AVX512DQ-FCP-NEXT: vmovdqa64 64(%rax), %zmm0
; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm23 = [0,5,4,0,0,6,5,0,0,5,4,0,0,6,5,0]
-; AVX512DQ-FCP-NEXT: # zmm23 = mem[0,1,2,3,0,1,2,3]
-; AVX512DQ-FCP-NEXT: vpermd %zmm0, %zmm23, %zmm0
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (zmm22 & (zmm0 ^ zmm3))
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm5))
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm2 = [0,0,4,0,0,0,5,0,0,5,0,0,0,6,0,0]
+; AVX512DQ-FCP-NEXT: vpermd %zmm0, %zmm2, %zmm0
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (zmm12 & (zmm0 ^ zmm3))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm4))
; AVX512DQ-FCP-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm8, %ymm1
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm0 = ymm18[3,3,3,3,7,7,7,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm12 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6,7,8],ymm1[9],ymm0[10,11],ymm1[12],ymm0[13,14,15]
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%rcx), %xmm5
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdx), %xmm6
-; AVX512DQ-FCP-NEXT: vpbroadcastq {{.*#+}} xmm4 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %xmm4, %xmm5, %xmm1
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm3 = xmm6[1,1,2,2]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2,3],xmm1[4],xmm3[5,6],xmm1[7]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm7, %ymm2
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm0 = ymm29[3,3,3,3,7,7,7,7]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm13 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6,7,8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14,15]
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%rcx), %xmm3
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdx), %xmm4
+; AVX512DQ-FCP-NEXT: vpbroadcastq {{.*#+}} xmm12 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vpshufb %xmm12, %xmm3, %xmm2
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm4, %xmm5
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[1,1,2,2]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0],xmm2[1],xmm4[2,3],xmm2[4],xmm4[5,6],xmm2[7]
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm14 = [0,1,2,3,0,1,2,3,6,7,4,5,6,7,4,5]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm14, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm22 = [8,9,9,0,0,0,1,1]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm22, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdi), %xmm6
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm6, (%rsp) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%rsi), %xmm7
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vprold $16, %xmm7, %xmm3
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm5 = xmm6[1,1,2,3]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm3 = xmm5[0,1],xmm3[2],xmm5[3,4],xmm3[5],xmm5[6,7]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm14, %xmm4, %xmm4
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm24 = [8,9,9,0,0,0,1,1]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm4, %zmm24, %zmm2
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%rdi), %xmm3
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm3, (%rsp) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%rsi), %xmm6
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vprold $16, %xmm6, %xmm4
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm5 = xmm3[1,1,2,3]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1],xmm4[2],xmm5[3,4],xmm4[5],xmm5[6,7]
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm15 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3]
; AVX512DQ-FCP-NEXT: vpshufb %xmm15, %xmm5, %xmm5
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm26 = [0,0,1,1,8,8,0,9]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm26, %zmm5
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm1 ^ (zmm3 & (zmm5 ^ zmm1))
-; AVX512DQ-FCP-NEXT: vmovdqa 64(%r9), %xmm1
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm4, %zmm26, %zmm5
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm4 = [65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm2 ^ (zmm4 & (zmm5 ^ zmm2))
+; AVX512DQ-FCP-NEXT: vmovdqa 64(%r9), %xmm2
; AVX512DQ-FCP-NEXT: vmovdqa 64(%r8), %xmm6
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm6[4],xmm1[4],xmm6[5],xmm1[5],xmm6[6],xmm1[6],xmm6[7],xmm1[7]
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm6[4],xmm2[4],xmm6[5],xmm2[5],xmm6[6],xmm2[6],xmm6[7],xmm2[7]
; AVX512DQ-FCP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm6[0],xmm1[0],xmm6[1],xmm1[1],xmm6[2],xmm1[2],xmm6[3],xmm1[3]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3]
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm7 = [0,1,2,3,6,7,4,5,6,7,4,5,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm1, %xmm6
-; AVX512DQ-FCP-NEXT: vpshufb %xmm7, %xmm1, %xmm1
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm18 = [0,0,0,1,8,9,9,0]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm18, %zmm1
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm2, %xmm6
+; AVX512DQ-FCP-NEXT: vpshufb %xmm7, %xmm2, %xmm2
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm29 = [0,0,0,1,8,9,9,0]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm29, %zmm2
; AVX512DQ-FCP-NEXT: vpbroadcastd 64(%rax), %ymm6
; AVX512DQ-FCP-NEXT: vpbroadcastd 68(%rax), %ymm8
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm6, %zmm29
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm30, %zmm11
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm29 = zmm29 ^ (zmm30 & (zmm29 ^ zmm1))
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm29 = zmm29 ^ (zmm1 & (zmm29 ^ zmm5))
-; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %xmm8
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm6, %zmm30
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm30 = zmm30 ^ (zmm20 & (zmm30 ^ zmm2))
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm30 = zmm30 ^ (zmm2 & (zmm30 ^ zmm5))
+; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %xmm3
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %xmm8
; AVX512DQ-FCP-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %xmm0
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpshufb %xmm4, %xmm8, %xmm5
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm6 = xmm0[1,1,2,2]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm12, %xmm3, %xmm5
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm6 = xmm8[1,1,2,2]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2,3],xmm5[4],xmm6[5,6],xmm5[7]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm3[0],xmm8[1],xmm3[1],xmm8[2],xmm3[2],xmm8[3],xmm3[3]
; AVX512DQ-FCP-NEXT: vpshufb %xmm14, %xmm6, %xmm6
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm22, %zmm5
-; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %xmm0
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %xmm4
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vprold $16, %xmm4, %xmm6
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm8 = xmm0[1,1,2,3]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm24, %zmm5
+; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %xmm3
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %xmm12
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vprold $16, %xmm12, %xmm6
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm8 = xmm3[1,1,2,3]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm6 = xmm8[0,1],xmm6[2],xmm8[3,4],xmm6[5],xmm8[6,7]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm3[0],xmm12[0],xmm3[1],xmm12[1],xmm3[2],xmm12[2],xmm3[3],xmm12[3]
; AVX512DQ-FCP-NEXT: vpshufb %xmm15, %xmm8, %xmm8
; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm26, %zmm8
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm5 ^ (zmm3 & (zmm8 ^ zmm5))
-; AVX512DQ-FCP-NEXT: vmovdqa (%r9), %xmm3
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm26, %zmm24
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm5 ^ (zmm4 & (zmm8 ^ zmm5))
+; AVX512DQ-FCP-NEXT: vmovdqa (%r9), %xmm4
; AVX512DQ-FCP-NEXT: vmovdqa (%r8), %xmm5
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
-; AVX512DQ-FCP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm2, %xmm3, %xmm5
-; AVX512DQ-FCP-NEXT: vpshufb %xmm7, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm7, %xmm24
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm18, %zmm3
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm0, %xmm4, %xmm5
+; AVX512DQ-FCP-NEXT: vpshufb %xmm7, %xmm4, %xmm4
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm7, %xmm26
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm29, %zmm4
; AVX512DQ-FCP-NEXT: vpbroadcastd (%rax), %ymm5
; AVX512DQ-FCP-NEXT: vpbroadcastd 4(%rax), %ymm6
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm6, %zmm5, %zmm30
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm30 = zmm30 ^ (zmm11 & (zmm30 ^ zmm3))
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm30 = zmm30 ^ (zmm1 & (zmm30 ^ zmm8))
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm6, %zmm5, %zmm29
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm29 = zmm29 ^ (zmm20 & (zmm29 ^ zmm4))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm29 = zmm29 ^ (zmm2 & (zmm29 ^ zmm8))
; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm21, %ymm0
-; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm7, %ymm1
+; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm7, %ymm2
; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm21 # 32-byte Reload
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm21[2,2,2,2,6,6,6,6]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5,6,7,8],ymm1[9],ymm3[10,11],ymm1[12],ymm3[13,14,15]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm7, %ymm3
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm21[2,2,2,2,6,6,6,6]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm2[1],ymm4[2,3],ymm2[4],ymm4[5,6,7,8],ymm2[9],ymm4[10,11],ymm2[12],ymm4[13,14,15]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm7, %ymm4
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm21[1,1,1,1,5,5,5,5]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm5[2],ymm3[3,4],ymm5[5],ymm3[6,7,8,9],ymm5[10],ymm3[11,12],ymm5[13],ymm3[14,15]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm19, %zmm3
-; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm9, %ymm1
-; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm18 # 32-byte Reload
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm18[2,2,2,2,6,6,6,6]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm5 = ymm1[0,1],ymm5[2],ymm1[3,4],ymm5[5],ymm1[6,7,8,9],ymm5[10],ymm1[11,12],ymm5[13],ymm1[14,15]
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm9[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm6 = ymm18[0,1,1,3,4,5,5,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm6[0,1],ymm1[2],ymm6[3,4],ymm1[5],ymm6[6,7,8,9],ymm1[10],ymm6[11,12],ymm1[13],ymm6[14,15]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm13, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqa 96(%rcx), %xmm11
-; AVX512DQ-FCP-NEXT: vmovdqa 96(%rdx), %xmm13
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm13[0],xmm11[0],xmm13[1],xmm11[1],xmm13[2],xmm11[2],xmm13[3],xmm11[3]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7,8,9],ymm5[10],ymm4[11,12],ymm5[13],ymm4[14,15]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm2, %zmm19, %zmm4
+; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm3, %ymm2
+; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm20 # 32-byte Reload
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm20[2,2,2,2,6,6,6,6]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm5 = ymm2[0,1],ymm5[2],ymm2[3,4],ymm5[5],ymm2[6,7,8,9],ymm5[10],ymm2[11,12],ymm5[13],ymm2[14,15]
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm3[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm6 = ymm20[0,1,1,3,4,5,5,7]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm6[0,1],ymm2[2],ymm6[3,4],ymm2[5],ymm6[6,7,8,9],ymm2[10],ymm6[11,12],ymm2[13],ymm6[14,15]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm11, %zmm2
+; AVX512DQ-FCP-NEXT: vmovdqa 96(%rcx), %xmm12
+; AVX512DQ-FCP-NEXT: vmovdqa 96(%rdx), %xmm11
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm11[0],xmm12[0],xmm11[1],xmm12[1],xmm11[2],xmm12[2],xmm11[3],xmm12[3]
; AVX512DQ-FCP-NEXT: vpshufb %xmm14, %xmm5, %xmm5
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm6 = [0,2,2,3,8,9,9,0]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm6, %zmm12
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm9 = [0,2,2,3,8,9,9,0]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm9, %zmm13
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm10 = [u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm27, %ymm2
-; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm2, %ymm5
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm8 = ymm28[3,3,3,3,7,7,7,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm8 = ymm8[0,1,2],ymm5[3],ymm8[4,5],ymm5[6],ymm8[7,8,9,10],ymm5[11],ymm8[12,13],ymm5[14],ymm8[15]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm3 ^ (mem & (zmm1 ^ zmm3))
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm22, %ymm5
+; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm5, %ymm5
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm6 = ymm28[3,3,3,3,7,7,7,7]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm8 = ymm6[0,1,2],ymm5[3],ymm6[4,5],ymm5[6],ymm6[7,8,9,10],ymm5[11],ymm6[12,13],ymm5[14],ymm6[15]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm4 ^ (zmm1 & (zmm2 ^ zmm4))
; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX512DQ-FCP-NEXT: vprold $16, %ymm5, %ymm2
-; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm20 # 32-byte Reload
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm20[1,2,2,3,5,6,6,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7,8,9],ymm2[10],ymm3[11,12],ymm2[13],ymm3[14,15]
-; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm3 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm5, %ymm31
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm20[0,0,2,1,4,4,6,5]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3],ymm3[4,5],ymm5[6],ymm3[7,8,9,10],ymm5[11],ymm3[12,13],ymm5[14],ymm3[15]
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm0 = [2,2,3,3,10,9,11,10]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm2, %zmm0, %zmm3
-; AVX512DQ-FCP-NEXT: vmovdqa 96(%rsi), %xmm2
-; AVX512DQ-FCP-NEXT: vmovdqa 96(%rdi), %xmm4
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm15, %xmm5, %xmm5
+; AVX512DQ-FCP-NEXT: vprold $16, %ymm5, %ymm1
+; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm17 # 32-byte Reload
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm4 = ymm17[1,2,2,3,5,6,6,7]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm4[0,1],ymm1[2],ymm4[3,4],ymm1[5],ymm4[6,7,8,9],ymm1[10],ymm4[11,12],ymm1[13],ymm4[14,15]
+; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm4 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm5, %ymm28
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm17[0,0,2,1,4,4,6,5]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3],ymm4[4,5],ymm5[6],ymm4[7,8,9,10],ymm5[11],ymm4[12,13],ymm5[14],ymm4[15]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm27, %zmm4
+; AVX512DQ-FCP-NEXT: vmovdqa 96(%rsi), %xmm1
+; AVX512DQ-FCP-NEXT: vmovdqa 96(%rdi), %xmm5
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm15, %xmm6, %xmm6
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm19 = [2,1,3,3,8,8,9,9]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm19, %zmm8
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm17 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm12 ^ (zmm17 & (zmm8 ^ zmm12))
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm19, %zmm8
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm13 ^ (zmm23 & (zmm8 ^ zmm13))
; AVX512DQ-FCP-NEXT: vbroadcasti32x4 {{.*#+}} ymm22 = [26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
; AVX512DQ-FCP-NEXT: # ymm22 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm22, %ymm0
-; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm5, %ymm0
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm25[3,3,3,3,7,7,7,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm12 = ymm0[0,1],ymm5[2],ymm0[3,4],ymm5[5],ymm0[6,7,8,9],ymm5[10],ymm0[11,12],ymm5[13],ymm0[14,15]
-; AVX512DQ-FCP-NEXT: vpermd %zmm16, %zmm23, %zmm16
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm16 = zmm16 ^ (mem & (zmm16 ^ zmm3))
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm16 = zmm16 ^ (mem & (zmm16 ^ zmm1))
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%rcx), %xmm1
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm18, %ymm6
+; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm6, %ymm0
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm6 = ymm25[3,3,3,3,7,7,7,7]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm6 = ymm0[0,1],ymm6[2],ymm0[3,4],ymm6[5],ymm0[6,7,8,9],ymm6[10],ymm0[11,12],ymm6[13],ymm0[14,15]
+; AVX512DQ-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,5,4,0,0,6,5,0,0,5,4,0,0,6,5,0]
+; AVX512DQ-FCP-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512DQ-FCP-NEXT: vpermd %zmm16, %zmm0, %zmm16
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm16 = zmm16 ^ (mem & (zmm16 ^ zmm4))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm16 = zmm16 ^ (mem & (zmm16 ^ zmm2))
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%rcx), %xmm2
+; AVX512DQ-FCP-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %xmm0
; AVX512DQ-FCP-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm1, %xmm28
-; AVX512DQ-FCP-NEXT: vpshufb %xmm14, %xmm0, %xmm5
-; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm1 = ymm9[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm18[3,3,3,3,7,7,7,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5,6,7,8],ymm1[9],ymm3[10,11],ymm1[12],ymm3[13,14,15]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm6, %zmm0
-; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm7, %ymm1
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm14, %xmm0, %xmm4
+; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm2 = ymm3[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm20[3,3,3,3,7,7,7,7]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5,6,7,8],ymm2[9],ymm3[10,11],ymm2[12],ymm3[13,14,15]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm4, %zmm9, %zmm0
+; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm7, %ymm2
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm3 = ymm21[3,3,3,3,7,7,7,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm14 = ymm3[0,1,2],ymm1[3],ymm3[4,5],ymm1[6],ymm3[7,8,9,10],ymm1[11],ymm3[12,13],ymm1[14],ymm3[15]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm14 = ymm3[0,1,2],ymm2[3],ymm3[4,5],ymm2[6],ymm3[7,8,9,10],ymm2[11],ymm3[12,13],ymm2[14],ymm3[15]
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %xmm10
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rsi), %xmm9
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm15, %xmm1, %xmm1
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm15, %xmm25
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm19, %zmm14
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm15, %xmm2, %xmm2
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm15, %xmm27
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm2, %zmm19, %zmm14
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm21 = [2,2,2,3,8,8,8,9]
-; AVX512DQ-FCP-NEXT: vmovdqa 96(%r9), %xmm6
-; AVX512DQ-FCP-NEXT: vmovdqa 96(%r8), %xmm5
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm24, %xmm18
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm24, %xmm1
-; AVX512DQ-FCP-NEXT: vpshufb %xmm1, %xmm7, %xmm15
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm15, %zmm21, %zmm12
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm14 = zmm0 ^ (zmm17 & (zmm14 ^ zmm0))
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm15 = [6,7,3,3,7,7,6,7]
-; AVX512DQ-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: vmovdqa 96(%r9), %xmm4
+; AVX512DQ-FCP-NEXT: vmovdqa 96(%r8), %xmm3
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm26, %xmm13
+; AVX512DQ-FCP-NEXT: vpshufb %xmm13, %xmm7, %xmm15
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm15, %zmm21, %zmm6
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm14 = zmm0 ^ (zmm23 & (zmm14 ^ zmm0))
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm18 = [6,7,3,3,7,7,6,7]
+; AVX512DQ-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm18, %ymm0 # 32-byte Folded Reload
; AVX512DQ-FCP-NEXT: vpbroadcastd 96(%rax), %ymm19
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm19, %zmm0, %zmm19
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm24 = [65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm19 = zmm19 ^ (zmm24 & (zmm19 ^ zmm12))
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm23 = [0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm19 = zmm19 ^ (zmm23 & (zmm19 ^ zmm8))
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
-; AVX512DQ-FCP-NEXT: vprold $16, %xmm2, %xmm0
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm2 = xmm4[1,1,2,3]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm23 = [65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm19 = zmm19 ^ (zmm23 & (zmm19 ^ zmm6))
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm26 = [0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm19 = zmm19 ^ (zmm26 & (zmm19 ^ zmm8))
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
+; AVX512DQ-FCP-NEXT: vprold $16, %xmm1, %xmm0
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm1 = xmm5[1,1,2,3]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3,4],xmm0[5],xmm1[6,7]
; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm22, %ymm1
-; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm31, %ymm2
-; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm2, %ymm1
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm2 = ymm20[3,3,3,3,7,7,7,7]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm28, %ymm5
+; AVX512DQ-FCP-NEXT: vpshufb %ymm1, %ymm5, %ymm1
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} ymm5 = ymm17[3,3,3,3,7,7,7,7]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm5[2],ymm1[3,4],ymm5[5],ymm1[6,7,8,9],ymm5[10],ymm1[11,12],ymm5[13],ymm1[14,15]
; AVX512DQ-FCP-NEXT: vmovdqa 32(%r9), %xmm8
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %xmm4
-; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm18, %xmm12
-; AVX512DQ-FCP-NEXT: vpshufb %xmm12, %xmm2, %xmm12
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm12, %zmm21, %zmm1
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm12 = [4,5,2,3,4,5,6,7,8,9,10,11,10,11,8,9]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm12, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm12, %xmm31
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm17 = [0,0,0,1,8,8,9,0]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm17, %zmm0
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm13[4],xmm11[4],xmm13[5],xmm11[5],xmm13[6],xmm11[6],xmm13[7],xmm11[7]
-; AVX512DQ-FCP-NEXT: vpbroadcastq {{.*#+}} xmm22 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm22, %xmm12
-; AVX512DQ-FCP-NEXT: vpshufb %xmm12, %xmm11, %xmm11
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[1,1,2,2]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm11 = xmm13[0],xmm11[1],xmm13[2,3],xmm11[4],xmm13[5,6],xmm11[7]
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,2,3,3,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm26, %zmm21
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm26, %zmm11
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm26 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm11 = zmm0 ^ (zmm26 & (zmm11 ^ zmm0))
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
-; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm20 # 32-byte Reload
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} ymm20 = ymm20 ^ (ymm24 & (ymm20 ^ ymm0))
-; AVX512DQ-FCP-NEXT: vextracti64x4 $1, %zmm0, %ymm0
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} xmm18 = [0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm18, %xmm5
-; AVX512DQ-FCP-NEXT: vpshufb %xmm5, %xmm7, %xmm5
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm25, %xmm12
-; AVX512DQ-FCP-NEXT: vpshufb %xmm12, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm25 = [8,9,9,0,0,0,1,1]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm25, %zmm3
-; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX512DQ-FCP-NEXT: vpshufhw {{.*#+}} ymm5 = ymm6[0,1,2,3,5,5,7,6,8,9,10,11,13,13,15,14]
-; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm5 = ymm5[3,3,3,3]
-; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,24,25,22,23,22,23,u,u,u,u,u,u,u,u]
-; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm6 = ymm6[2,2,2,2]
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %xmm6
+; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm8[0],xmm6[1],xmm8[1],xmm6[2],xmm8[2],xmm6[3],xmm8[3]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm13, %xmm5, %xmm15
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm15, %zmm21, %zmm1
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm13 = [4,5,2,3,4,5,6,7,8,9,10,11,10,11,8,9]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm13, %xmm2, %xmm2
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm21 = [0,0,0,1,8,8,9,0]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm2, %zmm21, %zmm0
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm11[4],xmm12[4],xmm11[5],xmm12[5],xmm11[6],xmm12[6],xmm11[7],xmm12[7]
+; AVX512DQ-FCP-NEXT: vpbroadcastq {{.*#+}} xmm20 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm20, %xmm15
+; AVX512DQ-FCP-NEXT: vpshufb %xmm15, %xmm12, %xmm12
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[1,1,2,2]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm12 = xmm11[0],xmm12[1],xmm11[2,3],xmm12[4],xmm11[5,6],xmm12[7]
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,2,3,3,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm24, %zmm25
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm2, %zmm24, %zmm12
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm28 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm12 = zmm0 ^ (zmm28 & (zmm12 ^ zmm0))
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
+; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm22 # 32-byte Reload
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} ymm22 = ymm22 ^ (ymm23 & (ymm22 ^ ymm2))
+; AVX512DQ-FCP-NEXT: vextracti64x4 $1, %zmm2, %ymm2
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} xmm24 = [0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm24, %xmm3
+; AVX512DQ-FCP-NEXT: vpshufb %xmm3, %xmm7, %xmm3
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm27, %xmm15
+; AVX512DQ-FCP-NEXT: vpshufb %xmm15, %xmm0, %xmm0
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm27 = [8,9,9,0,0,0,1,1]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm27, %zmm0
+; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX512DQ-FCP-NEXT: vpshufhw {{.*#+}} ymm3 = ymm4[0,1,2,3,5,5,7,6,8,9,10,11,13,13,15,14]
+; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm3 = ymm3[3,3,3,3]
+; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,24,25,22,23,22,23,u,u,u,u,u,u,u,u]
+; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,2,2,2]
; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX512DQ-FCP-NEXT: vpshufb %xmm12, %xmm7, %xmm7
-; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm13 = ymm7[0,0,1,1]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm15, %xmm7, %xmm7
+; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm11 = ymm7[0,0,1,1]
; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX512DQ-FCP-NEXT: vpshufb %xmm12, %xmm7, %xmm7
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm12, %xmm27
-; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm12 = ymm7[0,0,1,1]
-; AVX512DQ-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm7 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: vpshufb %xmm15, %xmm7, %xmm7
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm15, %xmm31
+; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm17 = ymm7[0,0,1,1]
+; AVX512DQ-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm18, %ymm7 # 32-byte Folded Reload
; AVX512DQ-FCP-NEXT: vpbroadcastd 32(%rax), %ymm15
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm15, %zmm7, %zmm7
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (zmm24 & (zmm7 ^ zmm1))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (zmm23 & (zmm7 ^ zmm1))
; AVX512DQ-FCP-NEXT: vpbroadcastd 100(%rax), %ymm1
; AVX512DQ-FCP-NEXT: vpbroadcastd 104(%rax), %ymm15
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm15, %zmm1, %zmm15
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm24 = [65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm15 = zmm15 ^ (zmm24 & (zmm15 ^ zmm3))
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (zmm23 & (zmm7 ^ zmm14))
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm15 = zmm15 ^ (zmm1 & (zmm15 ^ zmm11))
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm18 = [65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm15 = zmm15 ^ (zmm18 & (zmm15 ^ zmm0))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (zmm26 & (zmm7 ^ zmm14))
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm15 = zmm15 ^ (zmm0 & (zmm15 ^ zmm12))
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
; AVX512DQ-FCP-NEXT: vprold $16, %xmm9, %xmm9
; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[1,1,2,3]
; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0,1],xmm9[2],xmm10[3,4],xmm9[5],xmm10[6,7]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm31, %xmm14
-; AVX512DQ-FCP-NEXT: vpshufb %xmm14, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm3, %zmm17, %zmm9
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm22, %xmm3
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm28, %xmm10
-; AVX512DQ-FCP-NEXT: vpshufb %xmm3, %xmm10, %xmm3
-; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm10 = xmm11[4],xmm10[4],xmm11[5],xmm10[5],xmm11[6],xmm10[6],xmm11[7],xmm10[7]
-; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[1,1,2,2]
-; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm3 = xmm11[0],xmm3[1],xmm11[2,3],xmm3[4],xmm11[5,6],xmm3[7]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm13, %xmm1, %xmm1
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm1, %zmm21, %zmm9
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm20, %xmm1
+; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX512DQ-FCP-NEXT: vpshufb %xmm1, %xmm10, %xmm1
+; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm10 = xmm12[4],xmm10[4],xmm12[5],xmm10[5],xmm12[6],xmm10[6],xmm12[7],xmm10[7]
+; AVX512DQ-FCP-NEXT: vpshufd {{.*#+}} xmm12 = xmm12[1,1,2,2]
+; AVX512DQ-FCP-NEXT: vpblendw {{.*#+}} xmm1 = xmm12[0],xmm1[1],xmm12[2,3],xmm1[4],xmm12[5,6],xmm1[7]
; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,2,3,3,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm10, %zmm21, %zmm3
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm3 = zmm9 ^ (zmm26 & (zmm3 ^ zmm9))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm18, %xmm9
-; AVX512DQ-FCP-NEXT: vpshufb %xmm9, %xmm2, %xmm2
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm27, %xmm8
-; AVX512DQ-FCP-NEXT: vpshufb %xmm8, %xmm4, %xmm4
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm2, %zmm25, %zmm4
-; AVX512DQ-FCP-NEXT: vpbroadcastd 36(%rax), %ymm2
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm10, %zmm25, %zmm1
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm9 ^ (zmm28 & (zmm1 ^ zmm9))
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm24, %xmm9
+; AVX512DQ-FCP-NEXT: vpshufb %xmm9, %xmm5, %xmm5
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm6[4],xmm8[4],xmm6[5],xmm8[5],xmm6[6],xmm8[6],xmm6[7],xmm8[7]
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm31, %xmm8
+; AVX512DQ-FCP-NEXT: vpshufb %xmm8, %xmm6, %xmm6
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm27, %zmm6
+; AVX512DQ-FCP-NEXT: vpbroadcastd 36(%rax), %ymm5
; AVX512DQ-FCP-NEXT: vpbroadcastd 40(%rax), %ymm8
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm2, %zmm2
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm24 & (zmm2 ^ zmm4))
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (zmm1 & (zmm2 ^ zmm3))
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
-; AVX512DQ-FCP-NEXT: vpternlogd {{.*#+}} ymm0 = mem ^ (ymm1 & (ymm0 ^ mem))
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} ymm5 = ymm5 ^ (ymm1 & (ymm5 ^ ymm0))
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm4 = [65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} ymm6 = ymm6 ^ (ymm4 & (ymm6 ^ ymm20))
-; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm6[0,1,2,3],zmm5[0,1,2,3]
-; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
-; AVX512DQ-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm3 # 64-byte Folded Reload
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm3 = zmm3 ^ (mem & (zmm3 ^ zmm0))
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm13, %zmm0 # 32-byte Folded Reload
-; AVX512DQ-FCP-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm0 # 64-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # zmm0 = (zmm0 & zmm4) | mem
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm4, %zmm5
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm12, %zmm4 # 32-byte Folded Reload
-; AVX512DQ-FCP-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm5, %zmm4 # 64-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # zmm4 = (zmm4 & zmm5) | mem
-; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX512DQ-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm5, %zmm5
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (zmm18 & (zmm5 ^ zmm6))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (zmm0 & (zmm5 ^ zmm1))
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
+; AVX512DQ-FCP-NEXT: vpternlogd {{.*#+}} ymm2 = mem ^ (ymm0 & (ymm2 ^ mem))
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} ymm3 = ymm3 ^ (ymm0 & (ymm3 ^ ymm2))
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm6 = [65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} ymm4 = ymm4 ^ (ymm6 & (ymm4 ^ ymm22))
+; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm1 = zmm4[0,1,2,3],zmm3[0,1,2,3]
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm2 = [0,13,0,0,0,14,0,0,14,0,0,0,15,0,0,15]
+; AVX512DQ-FCP-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm2 # 64-byte Folded Reload
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm1))
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm11, %zmm1 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm6, %zmm1 # 64-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # zmm1 = (zmm1 & zmm6) | mem
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm17, %zmm3 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: vpternlogq $234, {{[-0-9]+}}(%r{{[sb]}}p), %zmm6, %zmm3 # 64-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # zmm3 = (zmm3 & zmm6) | mem
+; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX512DQ-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX512DQ-FCP-NEXT: vpunpckhwd (%rsp), %xmm6, %xmm6 # 16-byte Folded Reload
; AVX512DQ-FCP-NEXT: # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
@@ -15123,9 +15106,8 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX512DQ-FCP-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
; AVX512DQ-FCP-NEXT: # xmm9 = xmm9[4],mem[4],xmm9[5],mem[5],xmm9[6],mem[6],xmm9[7],mem[7]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm31, %xmm10
-; AVX512DQ-FCP-NEXT: vpshufb %xmm10, %xmm6, %xmm6
-; AVX512DQ-FCP-NEXT: vpshufb %xmm10, %xmm9, %xmm9
+; AVX512DQ-FCP-NEXT: vpshufb %xmm13, %xmm6, %xmm6
+; AVX512DQ-FCP-NEXT: vpshufb %xmm13, %xmm9, %xmm9
; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX512DQ-FCP-NEXT: vpunpcklwd {{.*#+}} ymm10 = ymm10[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
; AVX512DQ-FCP-NEXT: vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
@@ -15143,8 +15125,8 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm14 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535]
; AVX512DQ-FCP-NEXT: vpandn %ymm13, %ymm14, %ymm13
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm13, %zmm12, %zmm12
-; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,2,3,3,4,5,6,7]
-; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,0,2,1]
+; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,2,3,3,4,5,6,7]
+; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,0,2,1]
; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,0,1,3]
; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,2,3,3,4,5,6,7]
; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm8 = ymm8[0,0,2,1]
@@ -15154,43 +15136,43 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm14 = zmm14 ^ (mem & (zmm14 ^ zmm13))
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm5, %zmm5 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm4 # 32-byte Folded Reload
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm6, %zmm6 # 32-byte Folded Reload
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm5 ^ (zmm26 & (zmm6 ^ zmm5))
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm5 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm4 ^ (zmm28 & (zmm6 ^ zmm4))
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm8, %zmm4 # 32-byte Folded Reload
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm8 # 32-byte Folded Reload
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm5 ^ (zmm26 & (zmm8 ^ zmm5))
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm5 = [0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm0 ^ (zmm5 & (zmm6 ^ zmm0))
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm4 ^ (zmm5 & (zmm8 ^ zmm4))
-; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm10, %zmm0, %zmm0
-; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm11, %zmm4, %zmm4
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm0 ^ (mem & (zmm4 ^ zmm0))
-; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
-; AVX512DQ-FCP-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # zmm0 = zmm0 | (zmm1 & mem)
-; AVX512DQ-FCP-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm12 # 64-byte Folded Reload
-; AVX512DQ-FCP-NEXT: # zmm12 = zmm12 | (zmm1 & mem)
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm12 = zmm12 ^ (mem & (zmm12 ^ zmm4))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm4 ^ (zmm28 & (zmm8 ^ zmm4))
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm4 = [0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm1 ^ (zmm4 & (zmm6 ^ zmm1))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm3 ^ (zmm4 & (zmm8 ^ zmm3))
+; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm10, %zmm1, %zmm1
+; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm11, %zmm3, %zmm3
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm3 = zmm1 ^ (mem & (zmm3 ^ zmm1))
+; AVX512DQ-FCP-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
+; AVX512DQ-FCP-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 # 64-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # zmm1 = zmm1 | (zmm0 & mem)
+; AVX512DQ-FCP-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm12 # 64-byte Folded Reload
+; AVX512DQ-FCP-NEXT: # zmm12 = zmm12 | (zmm0 & mem)
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm12 = zmm12 ^ (mem & (zmm12 ^ zmm3))
; AVX512DQ-FCP-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm12, 320(%rax)
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, 256(%rax)
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm5, 256(%rax)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm7, 192(%rax)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm16, 128(%rax)
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm30, (%rax)
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm29, 448(%rax)
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm29, (%rax)
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm30, 448(%rax)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm15, 704(%rax)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm19, 640(%rax)
-; AVX512DQ-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
-; AVX512DQ-FCP-NEXT: vmovaps %zmm1, 576(%rax)
+; AVX512DQ-FCP-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512DQ-FCP-NEXT: vmovaps %zmm0, 576(%rax)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm14, 384(%rax)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm8, 64(%rax)
; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm6, 512(%rax)
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm3, 832(%rax)
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm0, 768(%rax)
-; AVX512DQ-FCP-NEXT: addq $1544, %rsp # imm = 0x608
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, 832(%rax)
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm1, 768(%rax)
+; AVX512DQ-FCP-NEXT: addq $1480, %rsp # imm = 0x5C8
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll
index a9e25ef7caad8..86efcf9c57616 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll
@@ -4667,27 +4667,27 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-NEXT: vmovdqa 32(%rsi), %ymm11
; AVX512-NEXT: vmovdqa {{.*#+}} ymm1 = [128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128,128,128,128]
; AVX512-NEXT: vpshufb %ymm1, %ymm11, %ymm0
-; AVX512-NEXT: vmovdqa64 %ymm1, %ymm20
-; AVX512-NEXT: vmovdqa 32(%rdi), %ymm8
+; AVX512-NEXT: vmovdqa64 %ymm1, %ymm18
+; AVX512-NEXT: vmovdqa 32(%rdi), %ymm5
; AVX512-NEXT: vmovdqa {{.*#+}} ymm2 = [12,13,128,15,12,13,14,128,12,13,14,15,128,u,u,u,16,128,18,19,16,17,128,19,16,17,18,128,16,17,18,19]
-; AVX512-NEXT: vpshufb %ymm2, %ymm8, %ymm1
-; AVX512-NEXT: vmovdqa64 %ymm2, %ymm21
-; AVX512-NEXT: vporq %ymm0, %ymm1, %ymm17
-; AVX512-NEXT: vmovdqa 32(%rdi), %xmm10
+; AVX512-NEXT: vpshufb %ymm2, %ymm5, %ymm1
+; AVX512-NEXT: vmovdqa64 %ymm2, %ymm19
+; AVX512-NEXT: vporq %ymm0, %ymm1, %ymm20
+; AVX512-NEXT: vmovdqa 32(%rdi), %xmm12
; AVX512-NEXT: vmovdqa {{.*#+}} xmm1 = [8,128,u,7,128,9,128,u,128,u,10,128,12,128,u,11]
-; AVX512-NEXT: vpshufb %xmm1, %xmm10, %xmm0
+; AVX512-NEXT: vpshufb %xmm1, %xmm12, %xmm0
; AVX512-NEXT: vmovdqa64 %xmm1, %xmm28
; AVX512-NEXT: vmovdqa 32(%rsi), %xmm9
; AVX512-NEXT: vmovdqa {{.*#+}} xmm2 = [128,8,u,128,7,128,9,u,11,u,128,10,128,12,u,128]
; AVX512-NEXT: vpshufb %xmm2, %xmm9, %xmm1
; AVX512-NEXT: vmovdqa64 %xmm2, %xmm29
-; AVX512-NEXT: vporq %xmm0, %xmm1, %xmm18
+; AVX512-NEXT: vporq %xmm0, %xmm1, %xmm21
; AVX512-NEXT: vmovdqa 32(%rcx), %ymm15
-; AVX512-NEXT: vmovdqa {{.*#+}} ymm5 = [128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128,128]
-; AVX512-NEXT: vpshufb %ymm5, %ymm15, %ymm0
-; AVX512-NEXT: vmovdqa 32(%rdx), %ymm13
-; AVX512-NEXT: vmovdqa {{.*#+}} ymm3 = [u,u,12,13,128,u,u,u,14,128,u,u,14,15,128,u,u,u,16,128,u,u,16,17,128,u,u,u,18,128,u,u]
-; AVX512-NEXT: vpshufb %ymm3, %ymm13, %ymm1
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm3 = [128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128,128]
+; AVX512-NEXT: vpshufb %ymm3, %ymm15, %ymm0
+; AVX512-NEXT: vmovdqa 32(%rdx), %ymm14
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm10 = [u,u,12,13,128,u,u,u,14,128,u,u,14,15,128,u,u,u,16,128,u,u,16,17,128,u,u,u,18,128,u,u]
+; AVX512-NEXT: vpshufb %ymm10, %ymm14, %ymm1
; AVX512-NEXT: vporq %ymm0, %ymm1, %ymm22
; AVX512-NEXT: vmovdqa 32(%rcx), %xmm6
; AVX512-NEXT: vmovdqa {{.*#+}} xmm1 = [128,6,128,8,u,128,7,128,9,128,11,u,128,10,128,12]
@@ -4698,72 +4698,72 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-NEXT: vpshufb %xmm2, %xmm7, %xmm1
; AVX512-NEXT: vmovdqa64 %xmm2, %xmm31
; AVX512-NEXT: vporq %xmm0, %xmm1, %xmm23
-; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm12 = [128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128,128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128]
-; AVX512-NEXT: # ymm12 = mem[0,1,0,1]
-; AVX512-NEXT: vpshufb %ymm12, %ymm8, %ymm0
+; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm8 = [128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128,128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128]
+; AVX512-NEXT: # ymm8 = mem[0,1,0,1]
+; AVX512-NEXT: vpshufb %ymm8, %ymm5, %ymm0
; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm4 = [128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,128,128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,128]
; AVX512-NEXT: # ymm4 = mem[0,1,0,1]
-; AVX512-NEXT: vpshufb %ymm4, %ymm8, %ymm1
+; AVX512-NEXT: vpshufb %ymm4, %ymm5, %ymm1
; AVX512-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm24
-; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm8 = [27,0,128,26,128,28,0,128,0,128,29,128,31,0,128,30,27,0,128,26,128,28,0,128,0,128,29,128,31,0,128,30]
-; AVX512-NEXT: # ymm8 = mem[0,1,0,1]
-; AVX512-NEXT: vpshufb %ymm8, %ymm11, %ymm1
+; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [27,0,128,26,128,28,0,128,0,128,29,128,31,0,128,30,27,0,128,26,128,28,0,128,0,128,29,128,31,0,128,30]
+; AVX512-NEXT: # ymm5 = mem[0,1,0,1]
+; AVX512-NEXT: vpshufb %ymm5, %ymm11, %ymm1
; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [19,128,21,0,128,20,128,22,128,24,0,128,23,128,25,0,19,128,21,0,128,20,128,22,128,24,0,128,23,128,25,0]
; AVX512-NEXT: # ymm2 = mem[0,1,0,1]
; AVX512-NEXT: vpshufb %ymm2, %ymm11, %ymm11
; AVX512-NEXT: vinserti64x4 $1, %ymm1, %zmm11, %zmm26
; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm11 = [128,27,128,128,26,128,28,128,30,128,128,29,128,31,128,128,128,27,128,128,26,128,28,128,30,128,128,29,128,31,128,128]
; AVX512-NEXT: # ymm11 = mem[0,1,0,1]
-; AVX512-NEXT: vpshufb %ymm11, %ymm13, %ymm1
+; AVX512-NEXT: vpshufb %ymm11, %ymm14, %ymm1
; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [128,128,19,128,21,128,128,20,128,22,128,24,128,128,23,128,128,128,19,128,21,128,128,20,128,22,128,24,128,128,23,128]
; AVX512-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX512-NEXT: vpshufb %ymm0, %ymm15, %ymm14
-; AVX512-NEXT: vinserti64x4 $1, %ymm1, %zmm14, %zmm25
+; AVX512-NEXT: vpshufb %ymm0, %ymm15, %ymm13
+; AVX512-NEXT: vinserti64x4 $1, %ymm1, %zmm13, %zmm25
; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [25,128,27,0,128,26,128,28,128,30,0,128,29,128,31,0,25,128,27,0,128,26,128,28,128,30,0,128,29,128,31,0]
; AVX512-NEXT: # ymm1 = mem[0,1,0,1]
-; AVX512-NEXT: vpshufb %ymm1, %ymm15, %ymm14
+; AVX512-NEXT: vpshufb %ymm1, %ymm15, %ymm13
; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm15 = [18,19,128,21,128,21,20,128,22,128,24,128,22,23,128,25,18,19,128,21,128,21,20,128,22,128,24,128,22,23,128,25]
; AVX512-NEXT: # ymm15 = mem[0,1,0,1]
-; AVX512-NEXT: vpshufb %ymm15, %ymm13, %ymm13
-; AVX512-NEXT: vinserti64x4 $1, %ymm14, %zmm13, %zmm27
-; AVX512-NEXT: vmovdqa (%rcx), %ymm13
-; AVX512-NEXT: vpshufb %ymm5, %ymm13, %ymm5
-; AVX512-NEXT: vmovdqa (%rdx), %ymm14
+; AVX512-NEXT: vpshufb %ymm15, %ymm14, %ymm14
+; AVX512-NEXT: vinserti64x4 $1, %ymm13, %zmm14, %zmm27
+; AVX512-NEXT: vmovdqa (%rcx), %ymm14
; AVX512-NEXT: vpshufb %ymm3, %ymm14, %ymm3
-; AVX512-NEXT: vporq %ymm5, %ymm3, %ymm16
-; AVX512-NEXT: vpshufb %ymm0, %ymm13, %ymm0
-; AVX512-NEXT: vpshufb %ymm15, %ymm14, %ymm3
-; AVX512-NEXT: vporq %ymm0, %ymm3, %ymm19
+; AVX512-NEXT: vmovdqa (%rdx), %ymm13
+; AVX512-NEXT: vpshufb %ymm10, %ymm13, %ymm10
+; AVX512-NEXT: vporq %ymm3, %ymm10, %ymm16
+; AVX512-NEXT: vpshufb %ymm0, %ymm14, %ymm0
+; AVX512-NEXT: vpshufb %ymm15, %ymm13, %ymm3
+; AVX512-NEXT: vporq %ymm0, %ymm3, %ymm17
; AVX512-NEXT: vmovdqa (%rsi), %ymm3
-; AVX512-NEXT: vmovdqa64 %ymm20, %ymm0
+; AVX512-NEXT: vmovdqa64 %ymm18, %ymm0
; AVX512-NEXT: vpshufb %ymm0, %ymm3, %ymm0
-; AVX512-NEXT: vmovdqa (%rdi), %ymm15
-; AVX512-NEXT: vmovdqa64 %ymm21, %ymm5
-; AVX512-NEXT: vpshufb %ymm5, %ymm15, %ymm5
-; AVX512-NEXT: vpor %ymm0, %ymm5, %ymm5
-; AVX512-NEXT: vpshufb %ymm4, %ymm15, %ymm0
+; AVX512-NEXT: vmovdqa (%rdi), %ymm10
+; AVX512-NEXT: vmovdqa64 %ymm19, %ymm15
+; AVX512-NEXT: vpshufb %ymm15, %ymm10, %ymm15
+; AVX512-NEXT: vporq %ymm0, %ymm15, %ymm18
+; AVX512-NEXT: vpshufb %ymm4, %ymm10, %ymm0
; AVX512-NEXT: vpshufb %ymm2, %ymm3, %ymm2
-; AVX512-NEXT: vpor %ymm0, %ymm2, %ymm2
-; AVX512-NEXT: vpshufb %ymm11, %ymm14, %ymm0
-; AVX512-NEXT: vpshufb %ymm1, %ymm13, %ymm1
-; AVX512-NEXT: vporq %ymm0, %ymm1, %ymm20
-; AVX512-NEXT: vpshufb %ymm12, %ymm15, %ymm0
-; AVX512-NEXT: vpshufb %ymm8, %ymm3, %ymm1
-; AVX512-NEXT: vporq %ymm0, %ymm1, %ymm21
-; AVX512-NEXT: vmovdqa (%rdi), %xmm0
+; AVX512-NEXT: vporq %ymm0, %ymm2, %ymm19
+; AVX512-NEXT: vpshufb %ymm11, %ymm13, %ymm0
+; AVX512-NEXT: vpshufb %ymm1, %ymm14, %ymm1
+; AVX512-NEXT: vpor %ymm0, %ymm1, %ymm0
+; AVX512-NEXT: vpshufb %ymm8, %ymm10, %ymm1
+; AVX512-NEXT: vpshufb %ymm5, %ymm3, %ymm2
+; AVX512-NEXT: vpor %ymm1, %ymm2, %ymm1
+; AVX512-NEXT: vmovdqa (%rdi), %xmm5
; AVX512-NEXT: vmovdqa (%rsi), %xmm8
-; AVX512-NEXT: vmovdqa64 %xmm28, %xmm1
-; AVX512-NEXT: vpshufb %xmm1, %xmm0, %xmm1
+; AVX512-NEXT: vmovdqa64 %xmm28, %xmm2
+; AVX512-NEXT: vpshufb %xmm2, %xmm5, %xmm2
; AVX512-NEXT: vmovdqa64 %xmm29, %xmm3
; AVX512-NEXT: vpshufb %xmm3, %xmm8, %xmm3
-; AVX512-NEXT: vpor %xmm1, %xmm3, %xmm11
-; AVX512-NEXT: vmovdqa (%rcx), %xmm1
+; AVX512-NEXT: vpor %xmm2, %xmm3, %xmm10
+; AVX512-NEXT: vmovdqa (%rcx), %xmm2
; AVX512-NEXT: vmovdqa (%rdx), %xmm3
; AVX512-NEXT: vmovdqa64 %xmm30, %xmm4
-; AVX512-NEXT: vpshufb %xmm4, %xmm1, %xmm4
-; AVX512-NEXT: vmovdqa64 %xmm31, %xmm12
-; AVX512-NEXT: vpshufb %xmm12, %xmm3, %xmm12
-; AVX512-NEXT: vpor %xmm4, %xmm12, %xmm12
+; AVX512-NEXT: vpshufb %xmm4, %xmm2, %xmm4
+; AVX512-NEXT: vmovdqa64 %xmm31, %xmm11
+; AVX512-NEXT: vpshufb %xmm11, %xmm3, %xmm11
+; AVX512-NEXT: vpor %xmm4, %xmm11, %xmm11
; AVX512-NEXT: vmovdqa 32(%r8), %ymm4
; AVX512-NEXT: vmovdqa {{.*#+}} ymm13 = [12,128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128]
; AVX512-NEXT: vpshufb %ymm13, %ymm4, %ymm4
@@ -4775,66 +4775,67 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-NEXT: vmovdqa (%r8), %ymm14
; AVX512-NEXT: vpshufb %ymm13, %ymm14, %ymm13
; AVX512-NEXT: vpshufd {{.*#+}} ymm14 = ymm14[0,2,1,1,4,6,5,5]
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm28 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0]
; AVX512-NEXT: vpermq {{.*#+}} ymm14 = ymm14[2,3,3,2]
-; AVX512-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm14, %ymm14
+; AVX512-NEXT: vpandnq %ymm14, %ymm28, %ymm14
; AVX512-NEXT: vinserti64x4 $1, %ymm14, %zmm13, %zmm13
; AVX512-NEXT: vpunpcklbw {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3],xmm6[4],xmm7[4],xmm6[5],xmm7[5],xmm6[6],xmm7[6],xmm6[7],xmm7[7]
; AVX512-NEXT: vmovdqa {{.*#+}} xmm7 = [2,u,1,0,5,4,u,3,u,7,6,11,10,u,9,8]
; AVX512-NEXT: vpshufb %xmm7, %xmm6, %xmm6
; AVX512-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,0,1,1]
-; AVX512-NEXT: vpunpcklbw {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3],xmm10[4],xmm9[4],xmm10[5],xmm9[5],xmm10[6],xmm9[6],xmm10[7],xmm9[7]
-; AVX512-NEXT: vmovdqa {{.*#+}} xmm10 = [0,1,4,5,u,2,3,6,7,10,11,u,8,9,12,13]
-; AVX512-NEXT: vpshufb %xmm10, %xmm9, %xmm9
+; AVX512-NEXT: vpunpcklbw {{.*#+}} xmm9 = xmm12[0],xmm9[0],xmm12[1],xmm9[1],xmm12[2],xmm9[2],xmm12[3],xmm9[3],xmm12[4],xmm9[4],xmm12[5],xmm9[5],xmm12[6],xmm9[6],xmm12[7],xmm9[7]
+; AVX512-NEXT: vmovdqa {{.*#+}} xmm12 = [0,1,4,5,u,2,3,6,7,10,11,u,8,9,12,13]
+; AVX512-NEXT: vpshufb %xmm12, %xmm9, %xmm9
; AVX512-NEXT: vpermq {{.*#+}} ymm9 = ymm9[0,0,1,1]
-; AVX512-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
-; AVX512-NEXT: vpshufb %xmm10, %xmm0, %xmm0
-; AVX512-NEXT: vinserti32x4 $2, %xmm11, %zmm0, %zmm0
-; AVX512-NEXT: vpunpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
-; AVX512-NEXT: vpshufb %xmm7, %xmm1, %xmm1
-; AVX512-NEXT: vinserti32x4 $2, %xmm12, %zmm1, %zmm1
-; AVX512-NEXT: vpermq {{.*#+}} ymm3 = ymm18[0,0,1,1]
-; AVX512-NEXT: vinserti64x4 $1, %ymm17, %zmm3, %zmm3
+; AVX512-NEXT: vpunpcklbw {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1],xmm5[2],xmm8[2],xmm5[3],xmm8[3],xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]
+; AVX512-NEXT: vpshufb %xmm12, %xmm5, %xmm5
+; AVX512-NEXT: vinserti32x4 $2, %xmm10, %zmm5, %zmm5
+; AVX512-NEXT: vpunpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; AVX512-NEXT: vpshufb %xmm7, %xmm2, %xmm2
+; AVX512-NEXT: vinserti32x4 $2, %xmm11, %zmm2, %zmm2
+; AVX512-NEXT: vpermq {{.*#+}} ymm3 = ymm21[0,0,1,1]
+; AVX512-NEXT: vinserti64x4 $1, %ymm20, %zmm3, %zmm3
; AVX512-NEXT: vpermq {{.*#+}} ymm7 = ymm23[0,0,1,1]
; AVX512-NEXT: vinserti64x4 $1, %ymm22, %zmm7, %zmm7
; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm8 = [255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0]
; AVX512-NEXT: vpternlogq {{.*#+}} zmm7 = zmm3 ^ (zmm8 & (zmm7 ^ zmm3))
; AVX512-NEXT: vpternlogq {{.*#+}} zmm4 = zmm4 | (zmm7 & zmm15)
-; AVX512-NEXT: vpermq {{.*#+}} ymm3 = ymm19[2,2,3,3]
-; AVX512-NEXT: vinserti64x4 $1, %ymm3, %zmm16, %zmm3
-; AVX512-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,2,3,3]
-; AVX512-NEXT: vinserti64x4 $1, %ymm2, %zmm5, %zmm2
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm2 = zmm3 ^ (zmm8 & (zmm2 ^ zmm3))
; AVX512-NEXT: vpermq {{.*#+}} zmm3 = zmm24[2,2,3,3,6,6,7,7]
-; AVX512-NEXT: vpermq {{.*#+}} zmm5 = zmm26[2,2,3,3,6,6,7,7]
-; AVX512-NEXT: vporq %zmm3, %zmm5, %zmm3
-; AVX512-NEXT: vpermq {{.*#+}} zmm5 = zmm25[2,2,3,3,6,6,7,7]
-; AVX512-NEXT: vpermq {{.*#+}} zmm7 = zmm27[2,2,3,3,6,6,7,7]
-; AVX512-NEXT: vporq %zmm5, %zmm7, %zmm5
-; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm7 = [0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255]
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm5 = zmm3 ^ (zmm7 & (zmm5 ^ zmm3))
-; AVX512-NEXT: vpermq {{.*#+}} ymm3 = ymm20[2,2,3,3]
-; AVX512-NEXT: vinserti64x4 $1, %ymm6, %zmm3, %zmm3
-; AVX512-NEXT: vpermq {{.*#+}} ymm6 = ymm21[2,2,3,3]
-; AVX512-NEXT: vinserti64x4 $1, %ymm9, %zmm6, %zmm6
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm6 = zmm3 ^ (zmm7 & (zmm6 ^ zmm3))
-; AVX512-NEXT: vmovdqa64 (%r8), %zmm3
-; AVX512-NEXT: vpmovsxbd {{.*#+}} zmm7 = [12,0,13,13,13,13,0,14,14,14,14,0,15,15,15,15]
-; AVX512-NEXT: vpermd %zmm3, %zmm7, %zmm7
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (mem & (zmm7 ^ zmm5))
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm13 = zmm13 | (zmm2 & mem)
-; AVX512-NEXT: vpmovsxbd {{.*#+}} zmm2 = [6,6,6,0,7,7,7,7,0,8,8,8,8,0,9,9]
-; AVX512-NEXT: vpermd %zmm3, %zmm2, %zmm2
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm6))
-; AVX512-NEXT: vpermq {{.*#+}} zmm0 = zmm0[0,0,1,1,4,4,5,5]
-; AVX512-NEXT: vpermq {{.*#+}} zmm1 = zmm1[0,0,1,1,4,4,5,5]
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (mem & (zmm1 ^ zmm0))
-; AVX512-NEXT: vpmovsxbd {{.*#+}} zmm0 = [0,0,0,0,0,0,1,1,1,1,0,2,2,2,2,0]
-; AVX512-NEXT: vpermd %zmm3, %zmm0, %zmm0
+; AVX512-NEXT: vpermq {{.*#+}} zmm7 = zmm26[2,2,3,3,6,6,7,7]
+; AVX512-NEXT: vporq %zmm3, %zmm7, %zmm3
+; AVX512-NEXT: vmovdqa64 (%r8), %zmm7
+; AVX512-NEXT: vpermq {{.*#+}} zmm10 = zmm25[2,2,3,3,6,6,7,7]
+; AVX512-NEXT: vpermq {{.*#+}} zmm11 = zmm27[2,2,3,3,6,6,7,7]
+; AVX512-NEXT: vporq %zmm10, %zmm11, %zmm10
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm11 = [0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255]
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm10 = zmm3 ^ (zmm11 & (zmm10 ^ zmm3))
+; AVX512-NEXT: vpmovsxbd {{.*#+}} zmm3 = [12,0,13,13,13,13,0,14,14,14,14,0,15,15,15,15]
+; AVX512-NEXT: vpermd %zmm7, %zmm3, %zmm3
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm3 = zmm3 ^ (zmm28 & (zmm3 ^ zmm10))
+; AVX512-NEXT: vpermq {{.*#+}} ymm10 = ymm17[2,2,3,3]
+; AVX512-NEXT: vinserti64x4 $1, %ymm10, %zmm16, %zmm10
+; AVX512-NEXT: vpermq {{.*#+}} ymm12 = ymm19[2,2,3,3]
+; AVX512-NEXT: vinserti64x4 $1, %ymm12, %zmm18, %zmm12
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm12 = zmm10 ^ (zmm8 & (zmm12 ^ zmm10))
+; AVX512-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,3,3]
+; AVX512-NEXT: vinserti64x4 $1, %ymm6, %zmm0, %zmm0
+; AVX512-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,3,3]
+; AVX512-NEXT: vinserti64x4 $1, %ymm9, %zmm1, %zmm1
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm1 = zmm0 ^ (zmm11 & (zmm1 ^ zmm0))
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm13 = zmm13 | (zmm12 & mem)
+; AVX512-NEXT: vpmovsxbd {{.*#+}} zmm0 = [6,6,6,0,7,7,7,7,0,8,8,8,8,0,9,9]
+; AVX512-NEXT: vpermd %zmm7, %zmm0, %zmm0
; AVX512-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm1))
+; AVX512-NEXT: vpermq {{.*#+}} zmm1 = zmm5[0,0,1,1,4,4,5,5]
+; AVX512-NEXT: vpermq {{.*#+}} zmm2 = zmm2[0,0,1,1,4,4,5,5]
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm1))
+; AVX512-NEXT: vpmovsxbd {{.*#+}} zmm1 = [0,0,0,0,0,0,1,1,1,1,0,2,2,2,2,0]
+; AVX512-NEXT: vpermd %zmm7, %zmm1, %zmm1
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (mem & (zmm1 ^ zmm2))
; AVX512-NEXT: vmovdqa64 %zmm13, 64(%r9)
-; AVX512-NEXT: vmovdqa64 %zmm0, (%r9)
-; AVX512-NEXT: vmovdqa64 %zmm2, 128(%r9)
-; AVX512-NEXT: vmovdqa64 %zmm7, 256(%r9)
+; AVX512-NEXT: vmovdqa64 %zmm1, (%r9)
+; AVX512-NEXT: vmovdqa64 %zmm0, 128(%r9)
+; AVX512-NEXT: vmovdqa64 %zmm3, 256(%r9)
; AVX512-NEXT: vmovdqa64 %zmm4, 192(%r9)
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
@@ -4847,52 +4848,53 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %ymm3
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm13 = [12,13,128,15,12,13,14,128,12,13,14,15,128,u,u,u,16,128,18,19,16,17,128,19,16,17,18,128,16,17,18,19]
; AVX512-FCP-NEXT: vpshufb %ymm13, %ymm3, %ymm2
-; AVX512-FCP-NEXT: vporq %ymm0, %ymm2, %ymm20
+; AVX512-FCP-NEXT: vpor %ymm0, %ymm2, %ymm0
+; AVX512-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512-FCP-NEXT: vmovdqa 32(%rdi), %xmm1
; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm0 = xmm1[8],zero,xmm1[u,7],zero,xmm1[9],zero,xmm1[u],zero,xmm1[u,10],zero,xmm1[12],zero,xmm1[u,11]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm1, %xmm28
+; AVX512-FCP-NEXT: vmovdqa64 %xmm1, %xmm29
; AVX512-FCP-NEXT: vmovdqa 32(%rsi), %xmm1
; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm2 = zero,xmm1[8,u],zero,xmm1[7],zero,xmm1[9,u,11,u],zero,xmm1[10],zero,xmm1[12,u],zero
-; AVX512-FCP-NEXT: vmovdqa64 %xmm1, %xmm29
-; AVX512-FCP-NEXT: vporq %xmm0, %xmm2, %xmm21
+; AVX512-FCP-NEXT: vmovdqa64 %xmm1, %xmm30
+; AVX512-FCP-NEXT: vporq %xmm0, %xmm2, %xmm22
; AVX512-FCP-NEXT: vmovdqa 32(%rcx), %ymm2
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm0 = [128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128,128]
; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm2, %ymm6
; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %ymm5
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm4 = [u,u,12,13,128,u,u,u,14,128,u,u,14,15,128,u,u,u,16,128,u,u,16,17,128,u,u,u,18,128,u,u]
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm5, %ymm7
-; AVX512-FCP-NEXT: vporq %ymm6, %ymm7, %ymm22
+; AVX512-FCP-NEXT: vporq %ymm6, %ymm7, %ymm23
; AVX512-FCP-NEXT: vmovdqa 32(%rcx), %xmm1
; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm6 = zero,xmm1[6],zero,xmm1[8,u],zero,xmm1[7],zero,xmm1[9],zero,xmm1[11,u],zero,xmm1[10],zero,xmm1[12]
-; AVX512-FCP-NEXT: vmovdqa64 %xmm1, %xmm30
+; AVX512-FCP-NEXT: vmovdqa64 %xmm1, %xmm31
; AVX512-FCP-NEXT: vmovdqa 32(%rdx), %xmm1
; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm7 = xmm1[6],zero,xmm1[8],zero,xmm1[u,7],zero,xmm1[9],zero,xmm1[11],zero,xmm1[u,10],zero,xmm1[12],zero
-; AVX512-FCP-NEXT: vmovdqa64 %xmm1, %xmm31
-; AVX512-FCP-NEXT: vporq %xmm6, %xmm7, %xmm23
+; AVX512-FCP-NEXT: vmovdqa64 %xmm1, %xmm21
+; AVX512-FCP-NEXT: vporq %xmm6, %xmm7, %xmm24
; AVX512-FCP-NEXT: vmovdqa (%rcx), %ymm6
; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm6, %ymm7
; AVX512-FCP-NEXT: vmovdqa (%rdx), %ymm1
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm1, %ymm4
-; AVX512-FCP-NEXT: vporq %ymm7, %ymm4, %ymm24
+; AVX512-FCP-NEXT: vporq %ymm7, %ymm4, %ymm25
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm14 = [128,128,19,128,21,128,128,20,128,22,128,24,128,128,23,128,128,128,19,128,21,128,128,20,128,22,128,24,128,128,23,128]
; AVX512-FCP-NEXT: # ymm14 = mem[0,1,0,1]
; AVX512-FCP-NEXT: vpshufb %ymm14, %ymm6, %ymm4
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm12 = [18,19,128,21,128,21,20,128,22,128,24,128,22,23,128,25,18,19,128,21,128,21,20,128,22,128,24,128,22,23,128,25]
; AVX512-FCP-NEXT: # ymm12 = mem[0,1,0,1]
; AVX512-FCP-NEXT: vpshufb %ymm12, %ymm1, %ymm9
-; AVX512-FCP-NEXT: vporq %ymm4, %ymm9, %ymm25
+; AVX512-FCP-NEXT: vporq %ymm4, %ymm9, %ymm26
; AVX512-FCP-NEXT: vmovdqa (%rsi), %ymm9
; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm9, %ymm4
; AVX512-FCP-NEXT: vmovdqa (%rdi), %ymm10
; AVX512-FCP-NEXT: vpshufb %ymm13, %ymm10, %ymm13
-; AVX512-FCP-NEXT: vporq %ymm4, %ymm13, %ymm26
+; AVX512-FCP-NEXT: vporq %ymm4, %ymm13, %ymm27
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm11 = [128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,128,128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,128]
; AVX512-FCP-NEXT: # ymm11 = mem[0,1,0,1]
; AVX512-FCP-NEXT: vpshufb %ymm11, %ymm10, %ymm7
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm4 = [19,128,21,0,128,20,128,22,128,24,0,128,23,128,25,0,19,128,21,0,128,20,128,22,128,24,0,128,23,128,25,0]
; AVX512-FCP-NEXT: # ymm4 = mem[0,1,0,1]
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm9, %ymm13
-; AVX512-FCP-NEXT: vporq %ymm7, %ymm13, %ymm27
+; AVX512-FCP-NEXT: vporq %ymm7, %ymm13, %ymm28
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm13 = [128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128,128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128]
; AVX512-FCP-NEXT: # ymm13 = mem[0,1,0,1]
; AVX512-FCP-NEXT: vpshufb %ymm13, %ymm3, %ymm0
@@ -4915,89 +4917,90 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-FCP-NEXT: vporq %ymm2, %ymm3, %ymm18
; AVX512-FCP-NEXT: vpshufb %ymm0, %ymm1, %ymm0
; AVX512-FCP-NEXT: vpshufb %ymm4, %ymm6, %ymm1
+; AVX512-FCP-NEXT: vporq %ymm0, %ymm1, %ymm20
+; AVX512-FCP-NEXT: vpshufb %ymm13, %ymm10, %ymm0
+; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm9, %ymm1
; AVX512-FCP-NEXT: vpor %ymm0, %ymm1, %ymm0
-; AVX512-FCP-NEXT: vpshufb %ymm13, %ymm10, %ymm1
-; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm9, %ymm2
-; AVX512-FCP-NEXT: vpor %ymm1, %ymm2, %ymm1
-; AVX512-FCP-NEXT: vmovdqa (%rdi), %xmm2
-; AVX512-FCP-NEXT: vmovdqa (%rsi), %xmm3
-; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm4 = xmm2[8],zero,xmm2[u,7],zero,xmm2[9],zero,xmm2[u],zero,xmm2[u,10],zero,xmm2[12],zero,xmm2[u,11]
-; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm5 = zero,xmm3[8,u],zero,xmm3[7],zero,xmm3[9,u,11,u],zero,xmm3[10],zero,xmm3[12,u],zero
-; AVX512-FCP-NEXT: vpor %xmm4, %xmm5, %xmm6
-; AVX512-FCP-NEXT: vmovdqa (%rcx), %xmm4
-; AVX512-FCP-NEXT: vmovdqa (%rdx), %xmm5
-; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm7 = zero,xmm4[6],zero,xmm4[8,u],zero,xmm4[7],zero,xmm4[9],zero,xmm4[11,u],zero,xmm4[10],zero,xmm4[12]
-; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm8 = xmm5[6],zero,xmm5[8],zero,xmm5[u,7],zero,xmm5[9],zero,xmm5[11],zero,xmm5[u,10],zero,xmm5[12],zero
-; AVX512-FCP-NEXT: vpor %xmm7, %xmm8, %xmm7
-; AVX512-FCP-NEXT: vmovdqa 32(%r8), %ymm8
-; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm9 = [12,128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128]
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm10 = [1,1,2,2,2,2,2,2]
-; AVX512-FCP-NEXT: vpermd %ymm8, %ymm10, %ymm10
-; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm8, %ymm8
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm11 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
-; AVX512-FCP-NEXT: vpandn %ymm10, %ymm11, %ymm10
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm10, %zmm8
-; AVX512-FCP-NEXT: vmovdqa (%r8), %ymm10
-; AVX512-FCP-NEXT: vpshufb %ymm9, %ymm10, %ymm9
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm12 = [4,0,5,5,5,5,0,6]
-; AVX512-FCP-NEXT: vpermd %ymm10, %ymm12, %ymm10
-; AVX512-FCP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm10, %ymm10
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm10, %zmm9, %zmm9
-; AVX512-FCP-NEXT: vmovdqa64 %xmm30, %xmm10
-; AVX512-FCP-NEXT: vmovdqa64 %xmm31, %xmm12
-; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} xmm10 = xmm10[0],xmm12[0],xmm10[1],xmm12[1],xmm10[2],xmm12[2],xmm10[3],xmm12[3],xmm10[4],xmm12[4],xmm10[5],xmm12[5],xmm10[6],xmm12[6],xmm10[7],xmm12[7]
+; AVX512-FCP-NEXT: vmovdqa (%rdi), %xmm1
+; AVX512-FCP-NEXT: vmovdqa (%rsi), %xmm2
+; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm3 = xmm1[8],zero,xmm1[u,7],zero,xmm1[9],zero,xmm1[u],zero,xmm1[u,10],zero,xmm1[12],zero,xmm1[u,11]
+; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm4 = zero,xmm2[8,u],zero,xmm2[7],zero,xmm2[9,u,11,u],zero,xmm2[10],zero,xmm2[12,u],zero
+; AVX512-FCP-NEXT: vpor %xmm3, %xmm4, %xmm5
+; AVX512-FCP-NEXT: vmovdqa (%rcx), %xmm3
+; AVX512-FCP-NEXT: vmovdqa (%rdx), %xmm4
+; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm6 = zero,xmm3[6],zero,xmm3[8,u],zero,xmm3[7],zero,xmm3[9],zero,xmm3[11,u],zero,xmm3[10],zero,xmm3[12]
+; AVX512-FCP-NEXT: vpshufb {{.*#+}} xmm7 = xmm4[6],zero,xmm4[8],zero,xmm4[u,7],zero,xmm4[9],zero,xmm4[11],zero,xmm4[u,10],zero,xmm4[12],zero
+; AVX512-FCP-NEXT: vpor %xmm6, %xmm7, %xmm6
+; AVX512-FCP-NEXT: vmovdqa 32(%r8), %ymm7
+; AVX512-FCP-NEXT: vmovdqa {{.*#+}} ymm8 = [12,128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128]
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm9 = [1,1,2,2,2,2,2,2]
+; AVX512-FCP-NEXT: vpermd %ymm7, %ymm9, %ymm9
+; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm7, %ymm7
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm10 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
+; AVX512-FCP-NEXT: vpandn %ymm9, %ymm10, %ymm9
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm7, %zmm9, %zmm7
+; AVX512-FCP-NEXT: vmovdqa (%r8), %ymm9
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} ymm11 = [4,0,5,5,5,5,0,6]
+; AVX512-FCP-NEXT: vpermd %ymm9, %ymm11, %ymm11
+; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm9, %ymm8
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm9 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0]
+; AVX512-FCP-NEXT: vpandn %ymm11, %ymm9, %ymm11
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm11, %zmm8, %zmm8
+; AVX512-FCP-NEXT: vmovdqa64 %xmm31, %xmm11
+; AVX512-FCP-NEXT: vmovdqa64 %xmm21, %xmm12
+; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} xmm11 = xmm11[0],xmm12[0],xmm11[1],xmm12[1],xmm11[2],xmm12[2],xmm11[3],xmm12[3],xmm11[4],xmm12[4],xmm11[5],xmm12[5],xmm11[6],xmm12[6],xmm11[7],xmm12[7]
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm12 = [2,u,1,0,5,4,u,3,u,7,6,11,10,u,9,8]
-; AVX512-FCP-NEXT: vpshufb %xmm12, %xmm10, %xmm10
+; AVX512-FCP-NEXT: vpshufb %xmm12, %xmm11, %xmm11
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm13 = [2,2,3,3,8,8,9,9]
-; AVX512-FCP-NEXT: vpermt2q %zmm10, %zmm13, %zmm0
-; AVX512-FCP-NEXT: vmovdqa64 %xmm28, %xmm10
-; AVX512-FCP-NEXT: vmovdqa64 %xmm29, %xmm14
-; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} xmm10 = xmm10[0],xmm14[0],xmm10[1],xmm14[1],xmm10[2],xmm14[2],xmm10[3],xmm14[3],xmm10[4],xmm14[4],xmm10[5],xmm14[5],xmm10[6],xmm14[6],xmm10[7],xmm14[7]
+; AVX512-FCP-NEXT: vpermt2q %zmm11, %zmm13, %zmm20
+; AVX512-FCP-NEXT: vmovdqa64 %xmm29, %xmm11
+; AVX512-FCP-NEXT: vmovdqa64 %xmm30, %xmm14
+; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} xmm11 = xmm11[0],xmm14[0],xmm11[1],xmm14[1],xmm11[2],xmm14[2],xmm11[3],xmm14[3],xmm11[4],xmm14[4],xmm11[5],xmm14[5],xmm11[6],xmm14[6],xmm11[7],xmm14[7]
; AVX512-FCP-NEXT: vmovdqa {{.*#+}} xmm14 = [0,1,4,5,u,2,3,6,7,10,11,u,8,9,12,13]
-; AVX512-FCP-NEXT: vpshufb %xmm14, %xmm10, %xmm10
-; AVX512-FCP-NEXT: vpermt2q %zmm10, %zmm13, %zmm1
-; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
-; AVX512-FCP-NEXT: vpshufb %xmm14, %xmm2, %xmm2
-; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} xmm3 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3],xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
-; AVX512-FCP-NEXT: vpshufb %xmm12, %xmm3, %xmm3
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm4 = [0,0,1,1,8,8,9,9]
-; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm4, %zmm2
-; AVX512-FCP-NEXT: vpermt2q %zmm7, %zmm4, %zmm3
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm3 = zmm3 ^ (mem & (zmm3 ^ zmm2))
-; AVX512-FCP-NEXT: vmovdqa64 (%r8), %zmm2
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm4 = [0,0,0,0,0,0,1,1,1,1,0,2,2,2,2,0]
-; AVX512-FCP-NEXT: vpermd %zmm2, %zmm4, %zmm4
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm4 ^ (mem & (zmm4 ^ zmm3))
-; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm3 = ymm21[0,0,1,1]
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm20, %zmm3, %zmm3
-; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm5 = ymm23[0,0,1,1]
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm22, %zmm5, %zmm5
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm6 = [255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm3 ^ (zmm6 & (zmm5 ^ zmm3))
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm8 | (zmm5 & zmm11)
-; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm3 = ymm25[2,2,3,3]
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm3, %zmm24, %zmm3
-; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm5 = ymm27[2,2,3,3]
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm5, %zmm26, %zmm5
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm3 ^ (zmm6 & (zmm5 ^ zmm3))
-; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm3 = [2,2,3,3,10,10,11,11]
-; AVX512-FCP-NEXT: vpermt2q %zmm16, %zmm3, %zmm17
-; AVX512-FCP-NEXT: vpermt2q %zmm19, %zmm3, %zmm18
-; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm18 = zmm17 ^ (zmm3 & (zmm18 ^ zmm17))
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm0 ^ (zmm3 & (zmm1 ^ zmm0))
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm9 = zmm9 | (zmm5 & mem)
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm0 = [12,0,13,13,13,13,0,14,14,14,14,0,15,15,15,15]
-; AVX512-FCP-NEXT: vpermd %zmm2, %zmm0, %zmm0
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm18))
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm3 = [6,6,6,0,7,7,7,7,0,8,8,8,8,0,9,9]
-; AVX512-FCP-NEXT: vpermd %zmm2, %zmm3, %zmm2
+; AVX512-FCP-NEXT: vpshufb %xmm14, %xmm11, %xmm11
+; AVX512-FCP-NEXT: vpermt2q %zmm11, %zmm13, %zmm0
+; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; AVX512-FCP-NEXT: vpshufb %xmm14, %xmm1, %xmm1
+; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} xmm2 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; AVX512-FCP-NEXT: vpshufb %xmm12, %xmm2, %xmm2
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm3 = [0,0,1,1,8,8,9,9]
+; AVX512-FCP-NEXT: vpermt2q %zmm5, %zmm3, %zmm1
+; AVX512-FCP-NEXT: vpermt2q %zmm6, %zmm3, %zmm2
; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm1))
-; AVX512-FCP-NEXT: vmovdqa64 %zmm4, (%r9)
-; AVX512-FCP-NEXT: vmovdqa64 %zmm2, 128(%r9)
-; AVX512-FCP-NEXT: vmovdqa64 %zmm0, 256(%r9)
-; AVX512-FCP-NEXT: vmovdqa64 %zmm9, 64(%r9)
-; AVX512-FCP-NEXT: vmovdqa64 %zmm8, 192(%r9)
+; AVX512-FCP-NEXT: vmovdqa64 (%r8), %zmm1
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm3 = [0,0,0,0,0,0,1,1,1,1,0,2,2,2,2,0]
+; AVX512-FCP-NEXT: vpermd %zmm1, %zmm3, %zmm3
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm3 = zmm3 ^ (mem & (zmm3 ^ zmm2))
+; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm2 = ymm22[0,0,1,1]
+; AVX512-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm2 # 32-byte Folded Reload
+; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm4 = ymm24[0,0,1,1]
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm23, %zmm4, %zmm4
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm5 = [255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm2 ^ (zmm5 & (zmm4 ^ zmm2))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 | (zmm4 & zmm10)
+; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm2 = ymm26[2,2,3,3]
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm25, %zmm2
+; AVX512-FCP-NEXT: vpermq {{.*#+}} ymm4 = ymm28[2,2,3,3]
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm27, %zmm4
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm2 ^ (zmm5 & (zmm4 ^ zmm2))
+; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} zmm2 = [2,2,3,3,10,10,11,11]
+; AVX512-FCP-NEXT: vpermt2q %zmm16, %zmm2, %zmm17
+; AVX512-FCP-NEXT: vpermt2q %zmm19, %zmm2, %zmm18
+; AVX512-FCP-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm18 = zmm17 ^ (zmm2 & (zmm18 ^ zmm17))
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm5 = [12,0,13,13,13,13,0,14,14,14,14,0,15,15,15,15]
+; AVX512-FCP-NEXT: vpermd %zmm1, %zmm5, %zmm5
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (zmm9 & (zmm5 ^ zmm18))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm20 ^ (zmm2 & (zmm0 ^ zmm20))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm8 | (zmm4 & mem)
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm2 = [6,6,6,0,7,7,7,7,0,8,8,8,8,0,9,9]
+; AVX512-FCP-NEXT: vpermd %zmm1, %zmm2, %zmm1
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (mem & (zmm1 ^ zmm0))
+; AVX512-FCP-NEXT: vmovdqa64 %zmm3, (%r9)
+; AVX512-FCP-NEXT: vmovdqa64 %zmm1, 128(%r9)
+; AVX512-FCP-NEXT: vmovdqa64 %zmm5, 256(%r9)
+; AVX512-FCP-NEXT: vmovdqa64 %zmm8, 64(%r9)
+; AVX512-FCP-NEXT: vmovdqa64 %zmm7, 192(%r9)
; AVX512-FCP-NEXT: vzeroupper
; AVX512-FCP-NEXT: retq
;
@@ -5006,27 +5009,27 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-NEXT: vmovdqa 32(%rsi), %ymm11
; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm1 = [128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128,128,128,128]
; AVX512DQ-NEXT: vpshufb %ymm1, %ymm11, %ymm0
-; AVX512DQ-NEXT: vmovdqa64 %ymm1, %ymm20
-; AVX512DQ-NEXT: vmovdqa 32(%rdi), %ymm8
+; AVX512DQ-NEXT: vmovdqa64 %ymm1, %ymm18
+; AVX512DQ-NEXT: vmovdqa 32(%rdi), %ymm5
; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm2 = [12,13,128,15,12,13,14,128,12,13,14,15,128,u,u,u,16,128,18,19,16,17,128,19,16,17,18,128,16,17,18,19]
-; AVX512DQ-NEXT: vpshufb %ymm2, %ymm8, %ymm1
-; AVX512DQ-NEXT: vmovdqa64 %ymm2, %ymm21
-; AVX512DQ-NEXT: vporq %ymm0, %ymm1, %ymm17
-; AVX512DQ-NEXT: vmovdqa 32(%rdi), %xmm10
+; AVX512DQ-NEXT: vpshufb %ymm2, %ymm5, %ymm1
+; AVX512DQ-NEXT: vmovdqa64 %ymm2, %ymm19
+; AVX512DQ-NEXT: vporq %ymm0, %ymm1, %ymm20
+; AVX512DQ-NEXT: vmovdqa 32(%rdi), %xmm12
; AVX512DQ-NEXT: vmovdqa {{.*#+}} xmm1 = [8,128,u,7,128,9,128,u,128,u,10,128,12,128,u,11]
-; AVX512DQ-NEXT: vpshufb %xmm1, %xmm10, %xmm0
+; AVX512DQ-NEXT: vpshufb %xmm1, %xmm12, %xmm0
; AVX512DQ-NEXT: vmovdqa64 %xmm1, %xmm28
; AVX512DQ-NEXT: vmovdqa 32(%rsi), %xmm9
; AVX512DQ-NEXT: vmovdqa {{.*#+}} xmm2 = [128,8,u,128,7,128,9,u,11,u,128,10,128,12,u,128]
; AVX512DQ-NEXT: vpshufb %xmm2, %xmm9, %xmm1
; AVX512DQ-NEXT: vmovdqa64 %xmm2, %xmm29
-; AVX512DQ-NEXT: vporq %xmm0, %xmm1, %xmm18
+; AVX512DQ-NEXT: vporq %xmm0, %xmm1, %xmm21
; AVX512DQ-NEXT: vmovdqa 32(%rcx), %ymm15
-; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm5 = [128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128,128]
-; AVX512DQ-NEXT: vpshufb %ymm5, %ymm15, %ymm0
-; AVX512DQ-NEXT: vmovdqa 32(%rdx), %ymm13
-; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm3 = [u,u,12,13,128,u,u,u,14,128,u,u,14,15,128,u,u,u,16,128,u,u,16,17,128,u,u,u,18,128,u,u]
-; AVX512DQ-NEXT: vpshufb %ymm3, %ymm13, %ymm1
+; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm3 = [128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128,128]
+; AVX512DQ-NEXT: vpshufb %ymm3, %ymm15, %ymm0
+; AVX512DQ-NEXT: vmovdqa 32(%rdx), %ymm14
+; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm10 = [u,u,12,13,128,u,u,u,14,128,u,u,14,15,128,u,u,u,16,128,u,u,16,17,128,u,u,u,18,128,u,u]
+; AVX512DQ-NEXT: vpshufb %ymm10, %ymm14, %ymm1
; AVX512DQ-NEXT: vporq %ymm0, %ymm1, %ymm22
; AVX512DQ-NEXT: vmovdqa 32(%rcx), %xmm6
; AVX512DQ-NEXT: vmovdqa {{.*#+}} xmm1 = [128,6,128,8,u,128,7,128,9,128,11,u,128,10,128,12]
@@ -5037,72 +5040,72 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-NEXT: vpshufb %xmm2, %xmm7, %xmm1
; AVX512DQ-NEXT: vmovdqa64 %xmm2, %xmm31
; AVX512DQ-NEXT: vporq %xmm0, %xmm1, %xmm23
-; AVX512DQ-NEXT: vbroadcasti128 {{.*#+}} ymm12 = [128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128,128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128]
-; AVX512DQ-NEXT: # ymm12 = mem[0,1,0,1]
-; AVX512DQ-NEXT: vpshufb %ymm12, %ymm8, %ymm0
+; AVX512DQ-NEXT: vbroadcasti128 {{.*#+}} ymm8 = [128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128,128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128]
+; AVX512DQ-NEXT: # ymm8 = mem[0,1,0,1]
+; AVX512DQ-NEXT: vpshufb %ymm8, %ymm5, %ymm0
; AVX512DQ-NEXT: vbroadcasti128 {{.*#+}} ymm4 = [128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,128,128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,128]
; AVX512DQ-NEXT: # ymm4 = mem[0,1,0,1]
-; AVX512DQ-NEXT: vpshufb %ymm4, %ymm8, %ymm1
+; AVX512DQ-NEXT: vpshufb %ymm4, %ymm5, %ymm1
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm24
-; AVX512DQ-NEXT: vbroadcasti128 {{.*#+}} ymm8 = [27,0,128,26,128,28,0,128,0,128,29,128,31,0,128,30,27,0,128,26,128,28,0,128,0,128,29,128,31,0,128,30]
-; AVX512DQ-NEXT: # ymm8 = mem[0,1,0,1]
-; AVX512DQ-NEXT: vpshufb %ymm8, %ymm11, %ymm1
+; AVX512DQ-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [27,0,128,26,128,28,0,128,0,128,29,128,31,0,128,30,27,0,128,26,128,28,0,128,0,128,29,128,31,0,128,30]
+; AVX512DQ-NEXT: # ymm5 = mem[0,1,0,1]
+; AVX512DQ-NEXT: vpshufb %ymm5, %ymm11, %ymm1
; AVX512DQ-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [19,128,21,0,128,20,128,22,128,24,0,128,23,128,25,0,19,128,21,0,128,20,128,22,128,24,0,128,23,128,25,0]
; AVX512DQ-NEXT: # ymm2 = mem[0,1,0,1]
; AVX512DQ-NEXT: vpshufb %ymm2, %ymm11, %ymm11
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm1, %zmm11, %zmm26
; AVX512DQ-NEXT: vbroadcasti128 {{.*#+}} ymm11 = [128,27,128,128,26,128,28,128,30,128,128,29,128,31,128,128,128,27,128,128,26,128,28,128,30,128,128,29,128,31,128,128]
; AVX512DQ-NEXT: # ymm11 = mem[0,1,0,1]
-; AVX512DQ-NEXT: vpshufb %ymm11, %ymm13, %ymm1
+; AVX512DQ-NEXT: vpshufb %ymm11, %ymm14, %ymm1
; AVX512DQ-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [128,128,19,128,21,128,128,20,128,22,128,24,128,128,23,128,128,128,19,128,21,128,128,20,128,22,128,24,128,128,23,128]
; AVX512DQ-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX512DQ-NEXT: vpshufb %ymm0, %ymm15, %ymm14
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm1, %zmm14, %zmm25
+; AVX512DQ-NEXT: vpshufb %ymm0, %ymm15, %ymm13
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm1, %zmm13, %zmm25
; AVX512DQ-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [25,128,27,0,128,26,128,28,128,30,0,128,29,128,31,0,25,128,27,0,128,26,128,28,128,30,0,128,29,128,31,0]
; AVX512DQ-NEXT: # ymm1 = mem[0,1,0,1]
-; AVX512DQ-NEXT: vpshufb %ymm1, %ymm15, %ymm14
+; AVX512DQ-NEXT: vpshufb %ymm1, %ymm15, %ymm13
; AVX512DQ-NEXT: vbroadcasti128 {{.*#+}} ymm15 = [18,19,128,21,128,21,20,128,22,128,24,128,22,23,128,25,18,19,128,21,128,21,20,128,22,128,24,128,22,23,128,25]
; AVX512DQ-NEXT: # ymm15 = mem[0,1,0,1]
-; AVX512DQ-NEXT: vpshufb %ymm15, %ymm13, %ymm13
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm14, %zmm13, %zmm27
-; AVX512DQ-NEXT: vmovdqa (%rcx), %ymm13
-; AVX512DQ-NEXT: vpshufb %ymm5, %ymm13, %ymm5
-; AVX512DQ-NEXT: vmovdqa (%rdx), %ymm14
+; AVX512DQ-NEXT: vpshufb %ymm15, %ymm14, %ymm14
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm13, %zmm14, %zmm27
+; AVX512DQ-NEXT: vmovdqa (%rcx), %ymm14
; AVX512DQ-NEXT: vpshufb %ymm3, %ymm14, %ymm3
-; AVX512DQ-NEXT: vporq %ymm5, %ymm3, %ymm16
-; AVX512DQ-NEXT: vpshufb %ymm0, %ymm13, %ymm0
-; AVX512DQ-NEXT: vpshufb %ymm15, %ymm14, %ymm3
-; AVX512DQ-NEXT: vporq %ymm0, %ymm3, %ymm19
+; AVX512DQ-NEXT: vmovdqa (%rdx), %ymm13
+; AVX512DQ-NEXT: vpshufb %ymm10, %ymm13, %ymm10
+; AVX512DQ-NEXT: vporq %ymm3, %ymm10, %ymm16
+; AVX512DQ-NEXT: vpshufb %ymm0, %ymm14, %ymm0
+; AVX512DQ-NEXT: vpshufb %ymm15, %ymm13, %ymm3
+; AVX512DQ-NEXT: vporq %ymm0, %ymm3, %ymm17
; AVX512DQ-NEXT: vmovdqa (%rsi), %ymm3
-; AVX512DQ-NEXT: vmovdqa64 %ymm20, %ymm0
+; AVX512DQ-NEXT: vmovdqa64 %ymm18, %ymm0
; AVX512DQ-NEXT: vpshufb %ymm0, %ymm3, %ymm0
-; AVX512DQ-NEXT: vmovdqa (%rdi), %ymm15
-; AVX512DQ-NEXT: vmovdqa64 %ymm21, %ymm5
-; AVX512DQ-NEXT: vpshufb %ymm5, %ymm15, %ymm5
-; AVX512DQ-NEXT: vpor %ymm0, %ymm5, %ymm5
-; AVX512DQ-NEXT: vpshufb %ymm4, %ymm15, %ymm0
+; AVX512DQ-NEXT: vmovdqa (%rdi), %ymm10
+; AVX512DQ-NEXT: vmovdqa64 %ymm19, %ymm15
+; AVX512DQ-NEXT: vpshufb %ymm15, %ymm10, %ymm15
+; AVX512DQ-NEXT: vporq %ymm0, %ymm15, %ymm18
+; AVX512DQ-NEXT: vpshufb %ymm4, %ymm10, %ymm0
; AVX512DQ-NEXT: vpshufb %ymm2, %ymm3, %ymm2
-; AVX512DQ-NEXT: vpor %ymm0, %ymm2, %ymm2
-; AVX512DQ-NEXT: vpshufb %ymm11, %ymm14, %ymm0
-; AVX512DQ-NEXT: vpshufb %ymm1, %ymm13, %ymm1
-; AVX512DQ-NEXT: vporq %ymm0, %ymm1, %ymm20
-; AVX512DQ-NEXT: vpshufb %ymm12, %ymm15, %ymm0
-; AVX512DQ-NEXT: vpshufb %ymm8, %ymm3, %ymm1
-; AVX512DQ-NEXT: vporq %ymm0, %ymm1, %ymm21
-; AVX512DQ-NEXT: vmovdqa (%rdi), %xmm0
+; AVX512DQ-NEXT: vporq %ymm0, %ymm2, %ymm19
+; AVX512DQ-NEXT: vpshufb %ymm11, %ymm13, %ymm0
+; AVX512DQ-NEXT: vpshufb %ymm1, %ymm14, %ymm1
+; AVX512DQ-NEXT: vpor %ymm0, %ymm1, %ymm0
+; AVX512DQ-NEXT: vpshufb %ymm8, %ymm10, %ymm1
+; AVX512DQ-NEXT: vpshufb %ymm5, %ymm3, %ymm2
+; AVX512DQ-NEXT: vpor %ymm1, %ymm2, %ymm1
+; AVX512DQ-NEXT: vmovdqa (%rdi), %xmm5
; AVX512DQ-NEXT: vmovdqa (%rsi), %xmm8
-; AVX512DQ-NEXT: vmovdqa64 %xmm28, %xmm1
-; AVX512DQ-NEXT: vpshufb %xmm1, %xmm0, %xmm1
+; AVX512DQ-NEXT: vmovdqa64 %xmm28, %xmm2
+; AVX512DQ-NEXT: vpshufb %xmm2, %xmm5, %xmm2
; AVX512DQ-NEXT: vmovdqa64 %xmm29, %xmm3
; AVX512DQ-NEXT: vpshufb %xmm3, %xmm8, %xmm3
-; AVX512DQ-NEXT: vpor %xmm1, %xmm3, %xmm11
-; AVX512DQ-NEXT: vmovdqa (%rcx), %xmm1
+; AVX512DQ-NEXT: vpor %xmm2, %xmm3, %xmm10
+; AVX512DQ-NEXT: vmovdqa (%rcx), %xmm2
; AVX512DQ-NEXT: vmovdqa (%rdx), %xmm3
; AVX512DQ-NEXT: vmovdqa64 %xmm30, %xmm4
-; AVX512DQ-NEXT: vpshufb %xmm4, %xmm1, %xmm4
-; AVX512DQ-NEXT: vmovdqa64 %xmm31, %xmm12
-; AVX512DQ-NEXT: vpshufb %xmm12, %xmm3, %xmm12
-; AVX512DQ-NEXT: vpor %xmm4, %xmm12, %xmm12
+; AVX512DQ-NEXT: vpshufb %xmm4, %xmm2, %xmm4
+; AVX512DQ-NEXT: vmovdqa64 %xmm31, %xmm11
+; AVX512DQ-NEXT: vpshufb %xmm11, %xmm3, %xmm11
+; AVX512DQ-NEXT: vpor %xmm4, %xmm11, %xmm11
; AVX512DQ-NEXT: vmovdqa 32(%r8), %ymm4
; AVX512DQ-NEXT: vmovdqa {{.*#+}} ymm13 = [12,128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128]
; AVX512DQ-NEXT: vpshufb %ymm13, %ymm4, %ymm4
@@ -5114,66 +5117,67 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-NEXT: vmovdqa (%r8), %ymm14
; AVX512DQ-NEXT: vpshufb %ymm13, %ymm14, %ymm13
; AVX512DQ-NEXT: vpshufd {{.*#+}} ymm14 = ymm14[0,2,1,1,4,6,5,5]
+; AVX512DQ-NEXT: vmovdqa64 {{.*#+}} zmm28 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0]
; AVX512DQ-NEXT: vpermq {{.*#+}} ymm14 = ymm14[2,3,3,2]
-; AVX512DQ-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm14, %ymm14
+; AVX512DQ-NEXT: vpandnq %ymm14, %ymm28, %ymm14
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm14, %zmm13, %zmm13
; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3],xmm6[4],xmm7[4],xmm6[5],xmm7[5],xmm6[6],xmm7[6],xmm6[7],xmm7[7]
; AVX512DQ-NEXT: vmovdqa {{.*#+}} xmm7 = [2,u,1,0,5,4,u,3,u,7,6,11,10,u,9,8]
; AVX512DQ-NEXT: vpshufb %xmm7, %xmm6, %xmm6
; AVX512DQ-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,0,1,1]
-; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3],xmm10[4],xmm9[4],xmm10[5],xmm9[5],xmm10[6],xmm9[6],xmm10[7],xmm9[7]
-; AVX512DQ-NEXT: vmovdqa {{.*#+}} xmm10 = [0,1,4,5,u,2,3,6,7,10,11,u,8,9,12,13]
-; AVX512DQ-NEXT: vpshufb %xmm10, %xmm9, %xmm9
+; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} xmm9 = xmm12[0],xmm9[0],xmm12[1],xmm9[1],xmm12[2],xmm9[2],xmm12[3],xmm9[3],xmm12[4],xmm9[4],xmm12[5],xmm9[5],xmm12[6],xmm9[6],xmm12[7],xmm9[7]
+; AVX512DQ-NEXT: vmovdqa {{.*#+}} xmm12 = [0,1,4,5,u,2,3,6,7,10,11,u,8,9,12,13]
+; AVX512DQ-NEXT: vpshufb %xmm12, %xmm9, %xmm9
; AVX512DQ-NEXT: vpermq {{.*#+}} ymm9 = ymm9[0,0,1,1]
-; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
-; AVX512DQ-NEXT: vpshufb %xmm10, %xmm0, %xmm0
-; AVX512DQ-NEXT: vinserti32x4 $2, %xmm11, %zmm0, %zmm0
-; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
-; AVX512DQ-NEXT: vpshufb %xmm7, %xmm1, %xmm1
-; AVX512DQ-NEXT: vinserti32x4 $2, %xmm12, %zmm1, %zmm1
-; AVX512DQ-NEXT: vpermq {{.*#+}} ymm3 = ymm18[0,0,1,1]
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm17, %zmm3, %zmm3
+; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1],xmm5[2],xmm8[2],xmm5[3],xmm8[3],xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]
+; AVX512DQ-NEXT: vpshufb %xmm12, %xmm5, %xmm5
+; AVX512DQ-NEXT: vinserti32x4 $2, %xmm10, %zmm5, %zmm5
+; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; AVX512DQ-NEXT: vpshufb %xmm7, %xmm2, %xmm2
+; AVX512DQ-NEXT: vinserti32x4 $2, %xmm11, %zmm2, %zmm2
+; AVX512DQ-NEXT: vpermq {{.*#+}} ymm3 = ymm21[0,0,1,1]
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm20, %zmm3, %zmm3
; AVX512DQ-NEXT: vpermq {{.*#+}} ymm7 = ymm23[0,0,1,1]
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm22, %zmm7, %zmm7
; AVX512DQ-NEXT: vmovdqa64 {{.*#+}} zmm8 = [255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0]
; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm7 = zmm3 ^ (zmm8 & (zmm7 ^ zmm3))
; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm4 = zmm4 | (zmm7 & zmm15)
-; AVX512DQ-NEXT: vpermq {{.*#+}} ymm3 = ymm19[2,2,3,3]
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm3, %zmm16, %zmm3
-; AVX512DQ-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,2,3,3]
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm2, %zmm5, %zmm2
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm2 = zmm3 ^ (zmm8 & (zmm2 ^ zmm3))
; AVX512DQ-NEXT: vpermq {{.*#+}} zmm3 = zmm24[2,2,3,3,6,6,7,7]
-; AVX512DQ-NEXT: vpermq {{.*#+}} zmm5 = zmm26[2,2,3,3,6,6,7,7]
-; AVX512DQ-NEXT: vporq %zmm3, %zmm5, %zmm3
-; AVX512DQ-NEXT: vpermq {{.*#+}} zmm5 = zmm25[2,2,3,3,6,6,7,7]
-; AVX512DQ-NEXT: vpermq {{.*#+}} zmm7 = zmm27[2,2,3,3,6,6,7,7]
-; AVX512DQ-NEXT: vporq %zmm5, %zmm7, %zmm5
-; AVX512DQ-NEXT: vmovdqa64 {{.*#+}} zmm7 = [0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255]
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm5 = zmm3 ^ (zmm7 & (zmm5 ^ zmm3))
-; AVX512DQ-NEXT: vpermq {{.*#+}} ymm3 = ymm20[2,2,3,3]
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm6, %zmm3, %zmm3
-; AVX512DQ-NEXT: vpermq {{.*#+}} ymm6 = ymm21[2,2,3,3]
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm9, %zmm6, %zmm6
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm6 = zmm3 ^ (zmm7 & (zmm6 ^ zmm3))
-; AVX512DQ-NEXT: vmovdqa64 (%r8), %zmm3
-; AVX512DQ-NEXT: vpmovsxbd {{.*#+}} zmm7 = [12,0,13,13,13,13,0,14,14,14,14,0,15,15,15,15]
-; AVX512DQ-NEXT: vpermd %zmm3, %zmm7, %zmm7
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 ^ (mem & (zmm7 ^ zmm5))
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm13 = zmm13 | (zmm2 & mem)
-; AVX512DQ-NEXT: vpmovsxbd {{.*#+}} zmm2 = [6,6,6,0,7,7,7,7,0,8,8,8,8,0,9,9]
-; AVX512DQ-NEXT: vpermd %zmm3, %zmm2, %zmm2
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm6))
-; AVX512DQ-NEXT: vpermq {{.*#+}} zmm0 = zmm0[0,0,1,1,4,4,5,5]
-; AVX512DQ-NEXT: vpermq {{.*#+}} zmm1 = zmm1[0,0,1,1,4,4,5,5]
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (mem & (zmm1 ^ zmm0))
-; AVX512DQ-NEXT: vpmovsxbd {{.*#+}} zmm0 = [0,0,0,0,0,0,1,1,1,1,0,2,2,2,2,0]
-; AVX512DQ-NEXT: vpermd %zmm3, %zmm0, %zmm0
+; AVX512DQ-NEXT: vpermq {{.*#+}} zmm7 = zmm26[2,2,3,3,6,6,7,7]
+; AVX512DQ-NEXT: vporq %zmm3, %zmm7, %zmm3
+; AVX512DQ-NEXT: vmovdqa64 (%r8), %zmm7
+; AVX512DQ-NEXT: vpermq {{.*#+}} zmm10 = zmm25[2,2,3,3,6,6,7,7]
+; AVX512DQ-NEXT: vpermq {{.*#+}} zmm11 = zmm27[2,2,3,3,6,6,7,7]
+; AVX512DQ-NEXT: vporq %zmm10, %zmm11, %zmm10
+; AVX512DQ-NEXT: vmovdqa64 {{.*#+}} zmm11 = [0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255]
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm10 = zmm3 ^ (zmm11 & (zmm10 ^ zmm3))
+; AVX512DQ-NEXT: vpmovsxbd {{.*#+}} zmm3 = [12,0,13,13,13,13,0,14,14,14,14,0,15,15,15,15]
+; AVX512DQ-NEXT: vpermd %zmm7, %zmm3, %zmm3
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm3 = zmm3 ^ (zmm28 & (zmm3 ^ zmm10))
+; AVX512DQ-NEXT: vpermq {{.*#+}} ymm10 = ymm17[2,2,3,3]
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm10, %zmm16, %zmm10
+; AVX512DQ-NEXT: vpermq {{.*#+}} ymm12 = ymm19[2,2,3,3]
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm12, %zmm18, %zmm12
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm12 = zmm10 ^ (zmm8 & (zmm12 ^ zmm10))
+; AVX512DQ-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,3,3]
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm6, %zmm0, %zmm0
+; AVX512DQ-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,3,3]
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm9, %zmm1, %zmm1
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm1 = zmm0 ^ (zmm11 & (zmm1 ^ zmm0))
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm13 = zmm13 | (zmm12 & mem)
+; AVX512DQ-NEXT: vpmovsxbd {{.*#+}} zmm0 = [6,6,6,0,7,7,7,7,0,8,8,8,8,0,9,9]
+; AVX512DQ-NEXT: vpermd %zmm7, %zmm0, %zmm0
; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm1))
+; AVX512DQ-NEXT: vpermq {{.*#+}} zmm1 = zmm5[0,0,1,1,4,4,5,5]
+; AVX512DQ-NEXT: vpermq {{.*#+}} zmm2 = zmm2[0,0,1,1,4,4,5,5]
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm1))
+; AVX512DQ-NEXT: vpmovsxbd {{.*#+}} zmm1 = [0,0,0,0,0,0,1,1,1,1,0,2,2,2,2,0]
+; AVX512DQ-NEXT: vpermd %zmm7, %zmm1, %zmm1
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (mem & (zmm1 ^ zmm2))
; AVX512DQ-NEXT: vmovdqa64 %zmm13, 64(%r9)
-; AVX512DQ-NEXT: vmovdqa64 %zmm0, (%r9)
-; AVX512DQ-NEXT: vmovdqa64 %zmm2, 128(%r9)
-; AVX512DQ-NEXT: vmovdqa64 %zmm7, 256(%r9)
+; AVX512DQ-NEXT: vmovdqa64 %zmm1, (%r9)
+; AVX512DQ-NEXT: vmovdqa64 %zmm0, 128(%r9)
+; AVX512DQ-NEXT: vmovdqa64 %zmm3, 256(%r9)
; AVX512DQ-NEXT: vmovdqa64 %zmm4, 192(%r9)
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
@@ -5186,52 +5190,53 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %ymm3
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm13 = [12,13,128,15,12,13,14,128,12,13,14,15,128,u,u,u,16,128,18,19,16,17,128,19,16,17,18,128,16,17,18,19]
; AVX512DQ-FCP-NEXT: vpshufb %ymm13, %ymm3, %ymm2
-; AVX512DQ-FCP-NEXT: vporq %ymm0, %ymm2, %ymm20
+; AVX512DQ-FCP-NEXT: vpor %ymm0, %ymm2, %ymm0
+; AVX512DQ-FCP-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdi), %xmm1
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm0 = xmm1[8],zero,xmm1[u,7],zero,xmm1[9],zero,xmm1[u],zero,xmm1[u,10],zero,xmm1[12],zero,xmm1[u,11]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm1, %xmm28
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm1, %xmm29
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rsi), %xmm1
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm2 = zero,xmm1[8,u],zero,xmm1[7],zero,xmm1[9,u,11,u],zero,xmm1[10],zero,xmm1[12,u],zero
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm1, %xmm29
-; AVX512DQ-FCP-NEXT: vporq %xmm0, %xmm2, %xmm21
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm1, %xmm30
+; AVX512DQ-FCP-NEXT: vporq %xmm0, %xmm2, %xmm22
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rcx), %ymm2
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm0 = [128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128,128]
; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm2, %ymm6
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %ymm5
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm4 = [u,u,12,13,128,u,u,u,14,128,u,u,14,15,128,u,u,u,16,128,u,u,16,17,128,u,u,u,18,128,u,u]
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm5, %ymm7
-; AVX512DQ-FCP-NEXT: vporq %ymm6, %ymm7, %ymm22
+; AVX512DQ-FCP-NEXT: vporq %ymm6, %ymm7, %ymm23
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rcx), %xmm1
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm6 = zero,xmm1[6],zero,xmm1[8,u],zero,xmm1[7],zero,xmm1[9],zero,xmm1[11,u],zero,xmm1[10],zero,xmm1[12]
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm1, %xmm30
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm1, %xmm31
; AVX512DQ-FCP-NEXT: vmovdqa 32(%rdx), %xmm1
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm7 = xmm1[6],zero,xmm1[8],zero,xmm1[u,7],zero,xmm1[9],zero,xmm1[11],zero,xmm1[u,10],zero,xmm1[12],zero
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm1, %xmm31
-; AVX512DQ-FCP-NEXT: vporq %xmm6, %xmm7, %xmm23
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm1, %xmm21
+; AVX512DQ-FCP-NEXT: vporq %xmm6, %xmm7, %xmm24
; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %ymm6
; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm6, %ymm7
; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %ymm1
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm1, %ymm4
-; AVX512DQ-FCP-NEXT: vporq %ymm7, %ymm4, %ymm24
+; AVX512DQ-FCP-NEXT: vporq %ymm7, %ymm4, %ymm25
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm14 = [128,128,19,128,21,128,128,20,128,22,128,24,128,128,23,128,128,128,19,128,21,128,128,20,128,22,128,24,128,128,23,128]
; AVX512DQ-FCP-NEXT: # ymm14 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpshufb %ymm14, %ymm6, %ymm4
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm12 = [18,19,128,21,128,21,20,128,22,128,24,128,22,23,128,25,18,19,128,21,128,21,20,128,22,128,24,128,22,23,128,25]
; AVX512DQ-FCP-NEXT: # ymm12 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpshufb %ymm12, %ymm1, %ymm9
-; AVX512DQ-FCP-NEXT: vporq %ymm4, %ymm9, %ymm25
+; AVX512DQ-FCP-NEXT: vporq %ymm4, %ymm9, %ymm26
; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %ymm9
; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm9, %ymm4
; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %ymm10
; AVX512DQ-FCP-NEXT: vpshufb %ymm13, %ymm10, %ymm13
-; AVX512DQ-FCP-NEXT: vporq %ymm4, %ymm13, %ymm26
+; AVX512DQ-FCP-NEXT: vporq %ymm4, %ymm13, %ymm27
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm11 = [128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,128,128,21,128,128,20,128,22,128,24,128,128,23,128,25,128,128]
; AVX512DQ-FCP-NEXT: # ymm11 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpshufb %ymm11, %ymm10, %ymm7
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm4 = [19,128,21,0,128,20,128,22,128,24,0,128,23,128,25,0,19,128,21,0,128,20,128,22,128,24,0,128,23,128,25,0]
; AVX512DQ-FCP-NEXT: # ymm4 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm9, %ymm13
-; AVX512DQ-FCP-NEXT: vporq %ymm7, %ymm13, %ymm27
+; AVX512DQ-FCP-NEXT: vporq %ymm7, %ymm13, %ymm28
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm13 = [128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128,128,128,26,128,28,128,128,27,128,29,128,31,128,128,30,128]
; AVX512DQ-FCP-NEXT: # ymm13 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpshufb %ymm13, %ymm3, %ymm0
@@ -5254,89 +5259,90 @@ define void @store_i8_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-FCP-NEXT: vporq %ymm2, %ymm3, %ymm18
; AVX512DQ-FCP-NEXT: vpshufb %ymm0, %ymm1, %ymm0
; AVX512DQ-FCP-NEXT: vpshufb %ymm4, %ymm6, %ymm1
+; AVX512DQ-FCP-NEXT: vporq %ymm0, %ymm1, %ymm20
+; AVX512DQ-FCP-NEXT: vpshufb %ymm13, %ymm10, %ymm0
+; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm9, %ymm1
; AVX512DQ-FCP-NEXT: vpor %ymm0, %ymm1, %ymm0
-; AVX512DQ-FCP-NEXT: vpshufb %ymm13, %ymm10, %ymm1
-; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm9, %ymm2
-; AVX512DQ-FCP-NEXT: vpor %ymm1, %ymm2, %ymm1
-; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %xmm2
-; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %xmm3
-; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm4 = xmm2[8],zero,xmm2[u,7],zero,xmm2[9],zero,xmm2[u],zero,xmm2[u,10],zero,xmm2[12],zero,xmm2[u,11]
-; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm5 = zero,xmm3[8,u],zero,xmm3[7],zero,xmm3[9,u,11,u],zero,xmm3[10],zero,xmm3[12,u],zero
-; AVX512DQ-FCP-NEXT: vpor %xmm4, %xmm5, %xmm6
-; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %xmm4
-; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %xmm5
-; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm7 = zero,xmm4[6],zero,xmm4[8,u],zero,xmm4[7],zero,xmm4[9],zero,xmm4[11,u],zero,xmm4[10],zero,xmm4[12]
-; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm8 = xmm5[6],zero,xmm5[8],zero,xmm5[u,7],zero,xmm5[9],zero,xmm5[11],zero,xmm5[u,10],zero,xmm5[12],zero
-; AVX512DQ-FCP-NEXT: vpor %xmm7, %xmm8, %xmm7
-; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %ymm8
-; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm9 = [12,128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128]
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm10 = [1,1,2,2,2,2,2,2]
-; AVX512DQ-FCP-NEXT: vpermd %ymm8, %ymm10, %ymm10
-; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm8, %ymm8
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm11 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
-; AVX512DQ-FCP-NEXT: vpandn %ymm10, %ymm11, %ymm10
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm10, %zmm8
-; AVX512DQ-FCP-NEXT: vmovdqa (%r8), %ymm10
-; AVX512DQ-FCP-NEXT: vpshufb %ymm9, %ymm10, %ymm9
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm12 = [4,0,5,5,5,5,0,6]
-; AVX512DQ-FCP-NEXT: vpermd %ymm10, %ymm12, %ymm10
-; AVX512DQ-FCP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm10, %ymm10
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm10, %zmm9, %zmm9
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm30, %xmm10
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm31, %xmm12
-; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} xmm10 = xmm10[0],xmm12[0],xmm10[1],xmm12[1],xmm10[2],xmm12[2],xmm10[3],xmm12[3],xmm10[4],xmm12[4],xmm10[5],xmm12[5],xmm10[6],xmm12[6],xmm10[7],xmm12[7]
+; AVX512DQ-FCP-NEXT: vmovdqa (%rdi), %xmm1
+; AVX512DQ-FCP-NEXT: vmovdqa (%rsi), %xmm2
+; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm3 = xmm1[8],zero,xmm1[u,7],zero,xmm1[9],zero,xmm1[u],zero,xmm1[u,10],zero,xmm1[12],zero,xmm1[u,11]
+; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm4 = zero,xmm2[8,u],zero,xmm2[7],zero,xmm2[9,u,11,u],zero,xmm2[10],zero,xmm2[12,u],zero
+; AVX512DQ-FCP-NEXT: vpor %xmm3, %xmm4, %xmm5
+; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %xmm3
+; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %xmm4
+; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm6 = zero,xmm3[6],zero,xmm3[8,u],zero,xmm3[7],zero,xmm3[9],zero,xmm3[11,u],zero,xmm3[10],zero,xmm3[12]
+; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} xmm7 = xmm4[6],zero,xmm4[8],zero,xmm4[u,7],zero,xmm4[9],zero,xmm4[11],zero,xmm4[u,10],zero,xmm4[12],zero
+; AVX512DQ-FCP-NEXT: vpor %xmm6, %xmm7, %xmm6
+; AVX512DQ-FCP-NEXT: vmovdqa 32(%r8), %ymm7
+; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} ymm8 = [12,128,128,128,128,13,128,128,128,128,14,128,128,128,128,15,128,128,128,128,16,128,128,128,128,17,128,128,128,128,18,128]
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm9 = [1,1,2,2,2,2,2,2]
+; AVX512DQ-FCP-NEXT: vpermd %ymm7, %ymm9, %ymm9
+; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm7, %ymm7
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm10 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
+; AVX512DQ-FCP-NEXT: vpandn %ymm9, %ymm10, %ymm9
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm7, %zmm9, %zmm7
+; AVX512DQ-FCP-NEXT: vmovdqa (%r8), %ymm9
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} ymm11 = [4,0,5,5,5,5,0,6]
+; AVX512DQ-FCP-NEXT: vpermd %ymm9, %ymm11, %ymm11
+; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm9, %ymm8
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm9 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255,0]
+; AVX512DQ-FCP-NEXT: vpandn %ymm11, %ymm9, %ymm11
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm11, %zmm8, %zmm8
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm31, %xmm11
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm21, %xmm12
+; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} xmm11 = xmm11[0],xmm12[0],xmm11[1],xmm12[1],xmm11[2],xmm12[2],xmm11[3],xmm12[3],xmm11[4],xmm12[4],xmm11[5],xmm12[5],xmm11[6],xmm12[6],xmm11[7],xmm12[7]
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm12 = [2,u,1,0,5,4,u,3,u,7,6,11,10,u,9,8]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm12, %xmm10, %xmm10
+; AVX512DQ-FCP-NEXT: vpshufb %xmm12, %xmm11, %xmm11
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm13 = [2,2,3,3,8,8,9,9]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm10, %zmm13, %zmm0
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm28, %xmm10
-; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm29, %xmm14
-; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} xmm10 = xmm10[0],xmm14[0],xmm10[1],xmm14[1],xmm10[2],xmm14[2],xmm10[3],xmm14[3],xmm10[4],xmm14[4],xmm10[5],xmm14[5],xmm10[6],xmm14[6],xmm10[7],xmm14[7]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm11, %zmm13, %zmm20
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm29, %xmm11
+; AVX512DQ-FCP-NEXT: vmovdqa64 %xmm30, %xmm14
+; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} xmm11 = xmm11[0],xmm14[0],xmm11[1],xmm14[1],xmm11[2],xmm14[2],xmm11[3],xmm14[3],xmm11[4],xmm14[4],xmm11[5],xmm14[5],xmm11[6],xmm14[6],xmm11[7],xmm14[7]
; AVX512DQ-FCP-NEXT: vmovdqa {{.*#+}} xmm14 = [0,1,4,5,u,2,3,6,7,10,11,u,8,9,12,13]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm14, %xmm10, %xmm10
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm10, %zmm13, %zmm1
-; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm14, %xmm2, %xmm2
-; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} xmm3 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3],xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
-; AVX512DQ-FCP-NEXT: vpshufb %xmm12, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm4 = [0,0,1,1,8,8,9,9]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm4, %zmm2
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm7, %zmm4, %zmm3
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm3 = zmm3 ^ (mem & (zmm3 ^ zmm2))
-; AVX512DQ-FCP-NEXT: vmovdqa64 (%r8), %zmm2
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm4 = [0,0,0,0,0,0,1,1,1,1,0,2,2,2,2,0]
-; AVX512DQ-FCP-NEXT: vpermd %zmm2, %zmm4, %zmm4
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm4 ^ (mem & (zmm4 ^ zmm3))
-; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm3 = ymm21[0,0,1,1]
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm20, %zmm3, %zmm3
-; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm5 = ymm23[0,0,1,1]
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm22, %zmm5, %zmm5
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm6 = [255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm3 ^ (zmm6 & (zmm5 ^ zmm3))
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm8 | (zmm5 & zmm11)
-; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm3 = ymm25[2,2,3,3]
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm3, %zmm24, %zmm3
-; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm5 = ymm27[2,2,3,3]
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm5, %zmm26, %zmm5
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm3 ^ (zmm6 & (zmm5 ^ zmm3))
-; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm3 = [2,2,3,3,10,10,11,11]
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm16, %zmm3, %zmm17
-; AVX512DQ-FCP-NEXT: vpermt2q %zmm19, %zmm3, %zmm18
-; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm18 = zmm17 ^ (zmm3 & (zmm18 ^ zmm17))
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm0 ^ (zmm3 & (zmm1 ^ zmm0))
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm9 = zmm9 | (zmm5 & mem)
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm0 = [12,0,13,13,13,13,0,14,14,14,14,0,15,15,15,15]
-; AVX512DQ-FCP-NEXT: vpermd %zmm2, %zmm0, %zmm0
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm18))
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm3 = [6,6,6,0,7,7,7,7,0,8,8,8,8,0,9,9]
-; AVX512DQ-FCP-NEXT: vpermd %zmm2, %zmm3, %zmm2
+; AVX512DQ-FCP-NEXT: vpshufb %xmm14, %xmm11, %xmm11
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm11, %zmm13, %zmm0
+; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm14, %xmm1, %xmm1
+; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} xmm2 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; AVX512DQ-FCP-NEXT: vpshufb %xmm12, %xmm2, %xmm2
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm3 = [0,0,1,1,8,8,9,9]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm5, %zmm3, %zmm1
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm6, %zmm3, %zmm2
; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm1))
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm4, (%r9)
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm2, 128(%r9)
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm0, 256(%r9)
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm9, 64(%r9)
-; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm8, 192(%r9)
+; AVX512DQ-FCP-NEXT: vmovdqa64 (%r8), %zmm1
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm3 = [0,0,0,0,0,0,1,1,1,1,0,2,2,2,2,0]
+; AVX512DQ-FCP-NEXT: vpermd %zmm1, %zmm3, %zmm3
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm3 = zmm3 ^ (mem & (zmm3 ^ zmm2))
+; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm2 = ymm22[0,0,1,1]
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm2 # 32-byte Folded Reload
+; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm4 = ymm24[0,0,1,1]
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm23, %zmm4, %zmm4
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm5 = [255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm2 ^ (zmm5 & (zmm4 ^ zmm2))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm7 = zmm7 | (zmm4 & zmm10)
+; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm2 = ymm26[2,2,3,3]
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm2, %zmm25, %zmm2
+; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} ymm4 = ymm28[2,2,3,3]
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm4, %zmm27, %zmm4
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm4 = zmm2 ^ (zmm5 & (zmm4 ^ zmm2))
+; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} zmm2 = [2,2,3,3,10,10,11,11]
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm16, %zmm2, %zmm17
+; AVX512DQ-FCP-NEXT: vpermt2q %zmm19, %zmm2, %zmm18
+; AVX512DQ-FCP-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255,255]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm18 = zmm17 ^ (zmm2 & (zmm18 ^ zmm17))
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm5 = [12,0,13,13,13,13,0,14,14,14,14,0,15,15,15,15]
+; AVX512DQ-FCP-NEXT: vpermd %zmm1, %zmm5, %zmm5
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm5 = zmm5 ^ (zmm9 & (zmm5 ^ zmm18))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm0 = zmm20 ^ (zmm2 & (zmm0 ^ zmm20))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm8 | (zmm4 & mem)
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm2 = [6,6,6,0,7,7,7,7,0,8,8,8,8,0,9,9]
+; AVX512DQ-FCP-NEXT: vpermd %zmm1, %zmm2, %zmm1
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm1 = zmm1 ^ (mem & (zmm1 ^ zmm0))
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm3, (%r9)
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm1, 128(%r9)
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm5, 256(%r9)
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm8, 64(%r9)
+; AVX512DQ-FCP-NEXT: vmovdqa64 %zmm7, 192(%r9)
; AVX512DQ-FCP-NEXT: vzeroupper
; AVX512DQ-FCP-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-6.ll
index 86e249649a291..887db41dcf8ad 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-6.ll
@@ -2608,10 +2608,10 @@ define void @store_i8_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-LABEL: store_i8_stride6_vf32:
; AVX512: # %bb.0:
; AVX512-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512-NEXT: vmovdqa (%rdi), %ymm3
-; AVX512-NEXT: vmovdqa (%rsi), %ymm5
-; AVX512-NEXT: vmovdqa (%rdx), %ymm2
-; AVX512-NEXT: vmovdqa (%rcx), %ymm4
+; AVX512-NEXT: vmovdqa (%rdi), %ymm2
+; AVX512-NEXT: vmovdqa (%rsi), %ymm3
+; AVX512-NEXT: vmovdqa (%rdx), %ymm4
+; AVX512-NEXT: vmovdqa (%rcx), %ymm5
; AVX512-NEXT: vmovdqa (%r8), %ymm0
; AVX512-NEXT: vmovdqa (%r9), %ymm1
; AVX512-NEXT: vmovdqa (%rcx), %xmm7
@@ -2619,7 +2619,7 @@ define void @store_i8_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm8[8],xmm7[8],xmm8[9],xmm7[9],xmm8[10],xmm7[10],xmm8[11],xmm7[11],xmm8[12],xmm7[12],xmm8[13],xmm7[13],xmm8[14],xmm7[14],xmm8[15],xmm7[15]
; AVX512-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[10,11,8,9,6,7,12,13,14,15,14,15,14,15,14,15]
; AVX512-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,0,0,1]
-; AVX512-NEXT: vpunpcklbw {{.*#+}} ymm9 = ymm2[0],ymm4[0],ymm2[1],ymm4[1],ymm2[2],ymm4[2],ymm2[3],ymm4[3],ymm2[4],ymm4[4],ymm2[5],ymm4[5],ymm2[6],ymm4[6],ymm2[7],ymm4[7],ymm2[16],ymm4[16],ymm2[17],ymm4[17],ymm2[18],ymm4[18],ymm2[19],ymm4[19],ymm2[20],ymm4[20],ymm2[21],ymm4[21],ymm2[22],ymm4[22],ymm2[23],ymm4[23]
+; AVX512-NEXT: vpunpcklbw {{.*#+}} ymm9 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[4],ymm5[4],ymm4[5],ymm5[5],ymm4[6],ymm5[6],ymm4[7],ymm5[7],ymm4[16],ymm5[16],ymm4[17],ymm5[17],ymm4[18],ymm5[18],ymm4[19],ymm5[19],ymm4[20],ymm5[20],ymm4[21],ymm5[21],ymm4[22],ymm5[22],ymm4[23],ymm5[23]
; AVX512-NEXT: vprold $16, %ymm9, %ymm9
; AVX512-NEXT: vpermq {{.*#+}} ymm9 = ymm9[2,2,2,3]
; AVX512-NEXT: vinserti64x4 $1, %ymm9, %zmm6, %zmm6
@@ -2628,7 +2628,7 @@ define void @store_i8_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-NEXT: vpunpckhbw {{.*#+}} xmm11 = xmm10[8],xmm9[8],xmm10[9],xmm9[9],xmm10[10],xmm9[10],xmm10[11],xmm9[11],xmm10[12],xmm9[12],xmm10[13],xmm9[13],xmm10[14],xmm9[14],xmm10[15],xmm9[15]
; AVX512-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[8,9,6,7,12,13,10,11,14,15,14,15,14,15,14,15]
; AVX512-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,0,0,1]
-; AVX512-NEXT: vpunpcklbw {{.*#+}} ymm12 = ymm3[0],ymm5[0],ymm3[1],ymm5[1],ymm3[2],ymm5[2],ymm3[3],ymm5[3],ymm3[4],ymm5[4],ymm3[5],ymm5[5],ymm3[6],ymm5[6],ymm3[7],ymm5[7],ymm3[16],ymm5[16],ymm3[17],ymm5[17],ymm3[18],ymm5[18],ymm3[19],ymm5[19],ymm3[20],ymm5[20],ymm3[21],ymm5[21],ymm3[22],ymm5[22],ymm3[23],ymm5[23]
+; AVX512-NEXT: vpunpcklbw {{.*#+}} ymm12 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[4],ymm3[4],ymm2[5],ymm3[5],ymm2[6],ymm3[6],ymm2[7],ymm3[7],ymm2[16],ymm3[16],ymm2[17],ymm3[17],ymm2[18],ymm3[18],ymm2[19],ymm3[19],ymm2[20],ymm3[20],ymm2[21],ymm3[21],ymm2[22],ymm3[22],ymm2[23],ymm3[23]
; AVX512-NEXT: vpshuflw {{.*#+}} ymm12 = ymm12[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
; AVX512-NEXT: vpshufhw {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
; AVX512-NEXT: vpermq {{.*#+}} ymm12 = ymm12[2,2,2,3]
@@ -2653,17 +2653,17 @@ define void @store_i8_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-NEXT: vpunpcklbw {{.*#+}} xmm15 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3],xmm8[4],xmm7[4],xmm8[5],xmm7[5],xmm8[6],xmm7[6],xmm8[7],xmm7[7]
; AVX512-NEXT: vprold $16, %xmm15, %xmm15
; AVX512-NEXT: vinserti64x4 $1, %ymm13, %zmm15, %zmm13
-; AVX512-NEXT: vpermq {{.*#+}} zmm13 = zmm13[0,0,0,1,4,4,4,5]
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm15 = [5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10]
-; AVX512-NEXT: vpshufb %xmm15, %xmm7, %xmm7
-; AVX512-NEXT: vpshufb %xmm15, %xmm8, %xmm8
+; AVX512-NEXT: vpermq {{.*#+}} zmm15 = zmm13[0,0,0,1,4,4,4,5]
+; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm13 = [5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10]
+; AVX512-NEXT: vpshufb %xmm13, %xmm7, %xmm7
+; AVX512-NEXT: vpshufb %xmm13, %xmm8, %xmm8
; AVX512-NEXT: vpunpcklbw {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3],xmm8[4],xmm7[4],xmm8[5],xmm7[5],xmm8[6],xmm7[6],xmm8[7],xmm7[7]
; AVX512-NEXT: vpunpcklbw {{.*#+}} xmm8 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3],xmm10[4],xmm9[4],xmm10[5],xmm9[5],xmm10[6],xmm9[6],xmm10[7],xmm9[7]
; AVX512-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,3,2,1,4,5,6,7]
; AVX512-NEXT: vpshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,6,5]
; AVX512-NEXT: vinserti64x4 $1, %ymm7, %zmm8, %zmm7
; AVX512-NEXT: vpermq {{.*#+}} zmm8 = zmm7[0,0,0,1,4,4,4,5]
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm8 = zmm13 ^ (mem & (zmm8 ^ zmm13))
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm8 = zmm15 ^ (mem & (zmm8 ^ zmm15))
; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm7 = [6,5,8,7,0,9,0,0,6,5,8,7,0,9,0,0,6,5,8,7,0,9,0,0,6,5,8,7,0,9,0,0]
; AVX512-NEXT: vpshufb %xmm7, %xmm11, %xmm9
; AVX512-NEXT: vpshufb %xmm7, %xmm12, %xmm10
@@ -2674,30 +2674,32 @@ define void @store_i8_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-NEXT: vinserti64x4 $1, %ymm9, %zmm10, %zmm9
; AVX512-NEXT: vpermq {{.*#+}} zmm9 = zmm9[0,0,0,1,4,4,4,5]
; AVX512-NEXT: vpternlogq {{.*#+}} zmm9 = zmm9 ^ (zmm14 & (zmm9 ^ zmm8))
-; AVX512-NEXT: vpshufb %ymm15, %ymm4, %ymm8
-; AVX512-NEXT: vpshufb %ymm15, %ymm2, %ymm10
-; AVX512-NEXT: vpunpcklbw {{.*#+}} ymm8 = ymm10[0],ymm8[0],ymm10[1],ymm8[1],ymm10[2],ymm8[2],ymm10[3],ymm8[3],ymm10[4],ymm8[4],ymm10[5],ymm8[5],ymm10[6],ymm8[6],ymm10[7],ymm8[7],ymm10[16],ymm8[16],ymm10[17],ymm8[17],ymm10[18],ymm8[18],ymm10[19],ymm8[19],ymm10[20],ymm8[20],ymm10[21],ymm8[21],ymm10[22],ymm8[22],ymm10[23],ymm8[23]
-; AVX512-NEXT: vpunpckhbw {{.*#+}} ymm10 = ymm3[8],ymm5[8],ymm3[9],ymm5[9],ymm3[10],ymm5[10],ymm3[11],ymm5[11],ymm3[12],ymm5[12],ymm3[13],ymm5[13],ymm3[14],ymm5[14],ymm3[15],ymm5[15],ymm3[24],ymm5[24],ymm3[25],ymm5[25],ymm3[26],ymm5[26],ymm3[27],ymm5[27],ymm3[28],ymm5[28],ymm3[29],ymm5[29],ymm3[30],ymm5[30],ymm3[31],ymm5[31]
-; AVX512-NEXT: vpshufb {{.*#+}} ymm10 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,22,23,28,29,26,27,30,31,30,31,30,31,30,31]
+; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm8 = [8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0]
+; AVX512-NEXT: vpshufb %ymm8, %ymm3, %ymm10
+; AVX512-NEXT: vpshufb %ymm8, %ymm2, %ymm8
+; AVX512-NEXT: vpunpcklbw {{.*#+}} ymm8 = ymm8[0],ymm10[0],ymm8[1],ymm10[1],ymm8[2],ymm10[2],ymm8[3],ymm10[3],ymm8[4],ymm10[4],ymm8[5],ymm10[5],ymm8[6],ymm10[6],ymm8[7],ymm10[7],ymm8[16],ymm10[16],ymm8[17],ymm10[17],ymm8[18],ymm10[18],ymm8[19],ymm10[19],ymm8[20],ymm10[20],ymm8[21],ymm10[21],ymm8[22],ymm10[22],ymm8[23],ymm10[23]
+; AVX512-NEXT: vpunpckhbw {{.*#+}} ymm10 = ymm4[8],ymm5[8],ymm4[9],ymm5[9],ymm4[10],ymm5[10],ymm4[11],ymm5[11],ymm4[12],ymm5[12],ymm4[13],ymm5[13],ymm4[14],ymm5[14],ymm4[15],ymm5[15],ymm4[24],ymm5[24],ymm4[25],ymm5[25],ymm4[26],ymm5[26],ymm4[27],ymm5[27],ymm4[28],ymm5[28],ymm4[29],ymm5[29],ymm4[30],ymm5[30],ymm4[31],ymm5[31]
+; AVX512-NEXT: vpshufb {{.*#+}} ymm10 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,22,23,28,29,30,31,30,31,30,31,30,31]
; AVX512-NEXT: vinserti64x4 $1, %ymm10, %zmm8, %zmm8
; AVX512-NEXT: vpermq {{.*#+}} zmm8 = zmm8[2,2,2,3,6,6,6,7]
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm10 = [8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0]
-; AVX512-NEXT: vpshufb %ymm10, %ymm5, %ymm5
-; AVX512-NEXT: vpshufb %ymm10, %ymm3, %ymm3
-; AVX512-NEXT: vpunpcklbw {{.*#+}} ymm3 = ymm3[0],ymm5[0],ymm3[1],ymm5[1],ymm3[2],ymm5[2],ymm3[3],ymm5[3],ymm3[4],ymm5[4],ymm3[5],ymm5[5],ymm3[6],ymm5[6],ymm3[7],ymm5[7],ymm3[16],ymm5[16],ymm3[17],ymm5[17],ymm3[18],ymm5[18],ymm3[19],ymm5[19],ymm3[20],ymm5[20],ymm3[21],ymm5[21],ymm3[22],ymm5[22],ymm3[23],ymm5[23]
-; AVX512-NEXT: vpunpckhbw {{.*#+}} ymm2 = ymm2[8],ymm4[8],ymm2[9],ymm4[9],ymm2[10],ymm4[10],ymm2[11],ymm4[11],ymm2[12],ymm4[12],ymm2[13],ymm4[13],ymm2[14],ymm4[14],ymm2[15],ymm4[15],ymm2[24],ymm4[24],ymm2[25],ymm4[25],ymm2[26],ymm4[26],ymm2[27],ymm4[27],ymm2[28],ymm4[28],ymm2[29],ymm4[29],ymm2[30],ymm4[30],ymm2[31],ymm4[31]
-; AVX512-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,22,23,28,29,30,31,30,31,30,31,30,31]
-; AVX512-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2
+; AVX512-NEXT: vpshufb %ymm13, %ymm5, %ymm5
+; AVX512-NEXT: vpshufb %ymm13, %ymm4, %ymm4
+; AVX512-NEXT: vpunpcklbw {{.*#+}} ymm4 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[4],ymm5[4],ymm4[5],ymm5[5],ymm4[6],ymm5[6],ymm4[7],ymm5[7],ymm4[16],ymm5[16],ymm4[17],ymm5[17],ymm4[18],ymm5[18],ymm4[19],ymm5[19],ymm4[20],ymm5[20],ymm4[21],ymm5[21],ymm4[22],ymm5[22],ymm4[23],ymm5[23]
+; AVX512-NEXT: vpunpckhbw {{.*#+}} ymm2 = ymm2[8],ymm3[8],ymm2[9],ymm3[9],ymm2[10],ymm3[10],ymm2[11],ymm3[11],ymm2[12],ymm3[12],ymm2[13],ymm3[13],ymm2[14],ymm3[14],ymm2[15],ymm3[15],ymm2[24],ymm3[24],ymm2[25],ymm3[25],ymm2[26],ymm3[26],ymm2[27],ymm3[27],ymm2[28],ymm3[28],ymm2[29],ymm3[29],ymm2[30],ymm3[30],ymm2[31],ymm3[31]
+; AVX512-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,22,23,28,29,26,27,30,31,30,31,30,31,30,31]
+; AVX512-NEXT: vinserti64x4 $1, %ymm2, %zmm4, %zmm2
; AVX512-NEXT: vpermq {{.*#+}} zmm2 = zmm2[2,2,2,3,6,6,6,7]
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm8))
-; AVX512-NEXT: vpshufb %ymm7, %ymm1, %ymm3
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm3 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512-NEXT: # zmm3 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm3 = zmm8 ^ (zmm3 & (zmm2 ^ zmm8))
+; AVX512-NEXT: vpshufb %ymm7, %ymm1, %ymm2
; AVX512-NEXT: vpshufb %ymm7, %ymm0, %ymm4
-; AVX512-NEXT: vpunpcklbw {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[4],ymm3[4],ymm4[5],ymm3[5],ymm4[6],ymm3[6],ymm4[7],ymm3[7],ymm4[16],ymm3[16],ymm4[17],ymm3[17],ymm4[18],ymm3[18],ymm4[19],ymm3[19],ymm4[20],ymm3[20],ymm4[21],ymm3[21],ymm4[22],ymm3[22],ymm4[23],ymm3[23]
+; AVX512-NEXT: vpunpcklbw {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[4],ymm2[4],ymm4[5],ymm2[5],ymm4[6],ymm2[6],ymm4[7],ymm2[7],ymm4[16],ymm2[16],ymm4[17],ymm2[17],ymm4[18],ymm2[18],ymm4[19],ymm2[19],ymm4[20],ymm2[20],ymm4[21],ymm2[21],ymm4[22],ymm2[22],ymm4[23],ymm2[23]
; AVX512-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
; AVX512-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27,24,25,22,23,28,29,26,27,28,29,30,31]
-; AVX512-NEXT: vinserti64x4 $1, %ymm0, %zmm3, %zmm0
+; AVX512-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm0
; AVX512-NEXT: vpermq {{.*#+}} zmm0 = zmm0[2,2,2,3,6,6,6,7]
-; AVX512-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm2))
+; AVX512-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm3))
; AVX512-NEXT: vmovdqa64 %zmm0, 128(%rax)
; AVX512-NEXT: vmovdqa64 %zmm9, (%rax)
; AVX512-NEXT: vmovdqa64 %zmm6, 64(%rax)
@@ -2713,32 +2715,34 @@ define void @store_i8_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-FCP-NEXT: vmovdqa (%rcx), %ymm5
; AVX512-FCP-NEXT: vmovdqa (%r8), %ymm0
; AVX512-FCP-NEXT: vmovdqa (%r9), %ymm1
-; AVX512-FCP-NEXT: vpbroadcastq {{.*#+}} ymm7 = [5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10]
-; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm5, %ymm6
-; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm4, %ymm8
-; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} ymm6 = ymm8[0],ymm6[0],ymm8[1],ymm6[1],ymm8[2],ymm6[2],ymm8[3],ymm6[3],ymm8[4],ymm6[4],ymm8[5],ymm6[5],ymm8[6],ymm6[6],ymm8[7],ymm6[7],ymm8[16],ymm6[16],ymm8[17],ymm6[17],ymm8[18],ymm6[18],ymm8[19],ymm6[19],ymm8[20],ymm6[20],ymm8[21],ymm6[21],ymm8[22],ymm6[22],ymm8[23],ymm6[23]
-; AVX512-FCP-NEXT: vpunpckhbw {{.*#+}} ymm8 = ymm2[8],ymm3[8],ymm2[9],ymm3[9],ymm2[10],ymm3[10],ymm2[11],ymm3[11],ymm2[12],ymm3[12],ymm2[13],ymm3[13],ymm2[14],ymm3[14],ymm2[15],ymm3[15],ymm2[24],ymm3[24],ymm2[25],ymm3[25],ymm2[26],ymm3[26],ymm2[27],ymm3[27],ymm2[28],ymm3[28],ymm2[29],ymm3[29],ymm2[30],ymm3[30],ymm2[31],ymm3[31]
-; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,22,23,28,29,26,27,30,31,30,31,30,31,30,31]
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm6, %zmm6
+; AVX512-FCP-NEXT: vpbroadcastq {{.*#+}} ymm6 = [8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0]
+; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm3, %ymm7
+; AVX512-FCP-NEXT: vpshufb %ymm6, %ymm2, %ymm6
+; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[1],ymm7[1],ymm6[2],ymm7[2],ymm6[3],ymm7[3],ymm6[4],ymm7[4],ymm6[5],ymm7[5],ymm6[6],ymm7[6],ymm6[7],ymm7[7],ymm6[16],ymm7[16],ymm6[17],ymm7[17],ymm6[18],ymm7[18],ymm6[19],ymm7[19],ymm6[20],ymm7[20],ymm6[21],ymm7[21],ymm6[22],ymm7[22],ymm6[23],ymm7[23]
+; AVX512-FCP-NEXT: vpunpckhbw {{.*#+}} ymm7 = ymm4[8],ymm5[8],ymm4[9],ymm5[9],ymm4[10],ymm5[10],ymm4[11],ymm5[11],ymm4[12],ymm5[12],ymm4[13],ymm5[13],ymm4[14],ymm5[14],ymm4[15],ymm5[15],ymm4[24],ymm5[24],ymm4[25],ymm5[25],ymm4[26],ymm5[26],ymm4[27],ymm5[27],ymm4[28],ymm5[28],ymm4[29],ymm5[29],ymm4[30],ymm5[30],ymm4[31],ymm5[31]
+; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,22,23,28,29,30,31,30,31,30,31,30,31]
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm7, %zmm6, %zmm6
; AVX512-FCP-NEXT: vpermq {{.*#+}} zmm6 = zmm6[2,2,2,3,6,6,6,7]
-; AVX512-FCP-NEXT: vpbroadcastq {{.*#+}} ymm8 = [8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0]
-; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm3, %ymm9
-; AVX512-FCP-NEXT: vpshufb %ymm8, %ymm2, %ymm8
-; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} ymm8 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[4],ymm9[4],ymm8[5],ymm9[5],ymm8[6],ymm9[6],ymm8[7],ymm9[7],ymm8[16],ymm9[16],ymm8[17],ymm9[17],ymm8[18],ymm9[18],ymm8[19],ymm9[19],ymm8[20],ymm9[20],ymm8[21],ymm9[21],ymm8[22],ymm9[22],ymm8[23],ymm9[23]
-; AVX512-FCP-NEXT: vpunpckhbw {{.*#+}} ymm9 = ymm4[8],ymm5[8],ymm4[9],ymm5[9],ymm4[10],ymm5[10],ymm4[11],ymm5[11],ymm4[12],ymm5[12],ymm4[13],ymm5[13],ymm4[14],ymm5[14],ymm4[15],ymm5[15],ymm4[24],ymm5[24],ymm4[25],ymm5[25],ymm4[26],ymm5[26],ymm4[27],ymm5[27],ymm4[28],ymm5[28],ymm4[29],ymm5[29],ymm4[30],ymm5[30],ymm4[31],ymm5[31]
-; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,22,23,28,29,30,31,30,31,30,31,30,31]
+; AVX512-FCP-NEXT: vpbroadcastq {{.*#+}} ymm7 = [5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10]
+; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm5, %ymm8
+; AVX512-FCP-NEXT: vpshufb %ymm7, %ymm4, %ymm9
+; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} ymm8 = ymm9[0],ymm8[0],ymm9[1],ymm8[1],ymm9[2],ymm8[2],ymm9[3],ymm8[3],ymm9[4],ymm8[4],ymm9[5],ymm8[5],ymm9[6],ymm8[6],ymm9[7],ymm8[7],ymm9[16],ymm8[16],ymm9[17],ymm8[17],ymm9[18],ymm8[18],ymm9[19],ymm8[19],ymm9[20],ymm8[20],ymm9[21],ymm8[21],ymm9[22],ymm8[22],ymm9[23],ymm8[23]
+; AVX512-FCP-NEXT: vpunpckhbw {{.*#+}} ymm9 = ymm2[8],ymm3[8],ymm2[9],ymm3[9],ymm2[10],ymm3[10],ymm2[11],ymm3[11],ymm2[12],ymm3[12],ymm2[13],ymm3[13],ymm2[14],ymm3[14],ymm2[15],ymm3[15],ymm2[24],ymm3[24],ymm2[25],ymm3[25],ymm2[26],ymm3[26],ymm2[27],ymm3[27],ymm2[28],ymm3[28],ymm2[29],ymm3[29],ymm2[30],ymm3[30],ymm2[31],ymm3[31]
+; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,22,23,28,29,26,27,30,31,30,31,30,31,30,31]
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm9, %zmm8, %zmm8
; AVX512-FCP-NEXT: vpermq {{.*#+}} zmm8 = zmm8[2,2,2,3,6,6,6,7]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm8 ^ (mem & (zmm8 ^ zmm6))
+; AVX512-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm9 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512-FCP-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3]
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm9 = zmm6 ^ (zmm9 & (zmm8 ^ zmm6))
; AVX512-FCP-NEXT: vpbroadcastq {{.*#+}} ymm10 = [6,5,8,7,0,9,0,0,6,5,8,7,0,9,0,0,6,5,8,7,0,9,0,0,6,5,8,7,0,9,0,0]
; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm1, %ymm6
-; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm0, %ymm9
-; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} ymm6 = ymm9[0],ymm6[0],ymm9[1],ymm6[1],ymm9[2],ymm6[2],ymm9[3],ymm6[3],ymm9[4],ymm6[4],ymm9[5],ymm6[5],ymm9[6],ymm6[6],ymm9[7],ymm6[7],ymm9[16],ymm6[16],ymm9[17],ymm6[17],ymm9[18],ymm6[18],ymm9[19],ymm6[19],ymm9[20],ymm6[20],ymm9[21],ymm6[21],ymm9[22],ymm6[22],ymm9[23],ymm6[23]
-; AVX512-FCP-NEXT: vpunpckhbw {{.*#+}} ymm9 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
-; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27,24,25,22,23,28,29,26,27,28,29,30,31]
-; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm9, %zmm6, %zmm6
+; AVX512-FCP-NEXT: vpshufb %ymm10, %ymm0, %ymm8
+; AVX512-FCP-NEXT: vpunpcklbw {{.*#+}} ymm6 = ymm8[0],ymm6[0],ymm8[1],ymm6[1],ymm8[2],ymm6[2],ymm8[3],ymm6[3],ymm8[4],ymm6[4],ymm8[5],ymm6[5],ymm8[6],ymm6[6],ymm8[7],ymm6[7],ymm8[16],ymm6[16],ymm8[17],ymm6[17],ymm8[18],ymm6[18],ymm8[19],ymm6[19],ymm8[20],ymm6[20],ymm8[21],ymm6[21],ymm8[22],ymm6[22],ymm8[23],ymm6[23]
+; AVX512-FCP-NEXT: vpunpckhbw {{.*#+}} ymm8 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
+; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27,24,25,22,23,28,29,26,27,28,29,30,31]
+; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm6, %zmm6
; AVX512-FCP-NEXT: vpermq {{.*#+}} zmm6 = zmm6[2,2,2,3,6,6,6,7]
-; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm6 ^ (mem & (zmm6 ^ zmm8))
+; AVX512-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm6 ^ (mem & (zmm6 ^ zmm9))
; AVX512-FCP-NEXT: vmovdqa (%rcx), %xmm9
; AVX512-FCP-NEXT: vpshufb %xmm7, %xmm9, %xmm8
; AVX512-FCP-NEXT: vmovdqa (%rdx), %xmm11
@@ -2797,10 +2801,10 @@ define void @store_i8_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-LABEL: store_i8_stride6_vf32:
; AVX512DQ: # %bb.0:
; AVX512DQ-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512DQ-NEXT: vmovdqa (%rdi), %ymm3
-; AVX512DQ-NEXT: vmovdqa (%rsi), %ymm5
-; AVX512DQ-NEXT: vmovdqa (%rdx), %ymm2
-; AVX512DQ-NEXT: vmovdqa (%rcx), %ymm4
+; AVX512DQ-NEXT: vmovdqa (%rdi), %ymm2
+; AVX512DQ-NEXT: vmovdqa (%rsi), %ymm3
+; AVX512DQ-NEXT: vmovdqa (%rdx), %ymm4
+; AVX512DQ-NEXT: vmovdqa (%rcx), %ymm5
; AVX512DQ-NEXT: vmovdqa (%r8), %ymm0
; AVX512DQ-NEXT: vmovdqa (%r9), %ymm1
; AVX512DQ-NEXT: vmovdqa (%rcx), %xmm7
@@ -2808,7 +2812,7 @@ define void @store_i8_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm8[8],xmm7[8],xmm8[9],xmm7[9],xmm8[10],xmm7[10],xmm8[11],xmm7[11],xmm8[12],xmm7[12],xmm8[13],xmm7[13],xmm8[14],xmm7[14],xmm8[15],xmm7[15]
; AVX512DQ-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[10,11,8,9,6,7,12,13,14,15,14,15,14,15,14,15]
; AVX512DQ-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,0,0,1]
-; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} ymm9 = ymm2[0],ymm4[0],ymm2[1],ymm4[1],ymm2[2],ymm4[2],ymm2[3],ymm4[3],ymm2[4],ymm4[4],ymm2[5],ymm4[5],ymm2[6],ymm4[6],ymm2[7],ymm4[7],ymm2[16],ymm4[16],ymm2[17],ymm4[17],ymm2[18],ymm4[18],ymm2[19],ymm4[19],ymm2[20],ymm4[20],ymm2[21],ymm4[21],ymm2[22],ymm4[22],ymm2[23],ymm4[23]
+; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} ymm9 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[4],ymm5[4],ymm4[5],ymm5[5],ymm4[6],ymm5[6],ymm4[7],ymm5[7],ymm4[16],ymm5[16],ymm4[17],ymm5[17],ymm4[18],ymm5[18],ymm4[19],ymm5[19],ymm4[20],ymm5[20],ymm4[21],ymm5[21],ymm4[22],ymm5[22],ymm4[23],ymm5[23]
; AVX512DQ-NEXT: vprold $16, %ymm9, %ymm9
; AVX512DQ-NEXT: vpermq {{.*#+}} ymm9 = ymm9[2,2,2,3]
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm9, %zmm6, %zmm6
@@ -2817,7 +2821,7 @@ define void @store_i8_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-NEXT: vpunpckhbw {{.*#+}} xmm11 = xmm10[8],xmm9[8],xmm10[9],xmm9[9],xmm10[10],xmm9[10],xmm10[11],xmm9[11],xmm10[12],xmm9[12],xmm10[13],xmm9[13],xmm10[14],xmm9[14],xmm10[15],xmm9[15]
; AVX512DQ-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[8,9,6,7,12,13,10,11,14,15,14,15,14,15,14,15]
; AVX512DQ-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,0,0,1]
-; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} ymm12 = ymm3[0],ymm5[0],ymm3[1],ymm5[1],ymm3[2],ymm5[2],ymm3[3],ymm5[3],ymm3[4],ymm5[4],ymm3[5],ymm5[5],ymm3[6],ymm5[6],ymm3[7],ymm5[7],ymm3[16],ymm5[16],ymm3[17],ymm5[17],ymm3[18],ymm5[18],ymm3[19],ymm5[19],ymm3[20],ymm5[20],ymm3[21],ymm5[21],ymm3[22],ymm5[22],ymm3[23],ymm5[23]
+; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} ymm12 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[4],ymm3[4],ymm2[5],ymm3[5],ymm2[6],ymm3[6],ymm2[7],ymm3[7],ymm2[16],ymm3[16],ymm2[17],ymm3[17],ymm2[18],ymm3[18],ymm2[19],ymm3[19],ymm2[20],ymm3[20],ymm2[21],ymm3[21],ymm2[22],ymm3[22],ymm2[23],ymm3[23]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} ymm12 = ymm12[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
; AVX512DQ-NEXT: vpshufhw {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
; AVX512DQ-NEXT: vpermq {{.*#+}} ymm12 = ymm12[2,2,2,3]
@@ -2842,17 +2846,17 @@ define void @store_i8_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} xmm15 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3],xmm8[4],xmm7[4],xmm8[5],xmm7[5],xmm8[6],xmm7[6],xmm8[7],xmm7[7]
; AVX512DQ-NEXT: vprold $16, %xmm15, %xmm15
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm13, %zmm15, %zmm13
-; AVX512DQ-NEXT: vpermq {{.*#+}} zmm13 = zmm13[0,0,0,1,4,4,4,5]
-; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm15 = [5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10]
-; AVX512DQ-NEXT: vpshufb %xmm15, %xmm7, %xmm7
-; AVX512DQ-NEXT: vpshufb %xmm15, %xmm8, %xmm8
+; AVX512DQ-NEXT: vpermq {{.*#+}} zmm15 = zmm13[0,0,0,1,4,4,4,5]
+; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm13 = [5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10]
+; AVX512DQ-NEXT: vpshufb %xmm13, %xmm7, %xmm7
+; AVX512DQ-NEXT: vpshufb %xmm13, %xmm8, %xmm8
; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3],xmm8[4],xmm7[4],xmm8[5],xmm7[5],xmm8[6],xmm7[6],xmm8[7],xmm7[7]
; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} xmm8 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3],xmm10[4],xmm9[4],xmm10[5],xmm9[5],xmm10[6],xmm9[6],xmm10[7],xmm9[7]
; AVX512DQ-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,3,2,1,4,5,6,7]
; AVX512DQ-NEXT: vpshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,6,5]
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm7, %zmm8, %zmm7
; AVX512DQ-NEXT: vpermq {{.*#+}} zmm8 = zmm7[0,0,0,1,4,4,4,5]
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm8 = zmm13 ^ (mem & (zmm8 ^ zmm13))
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm8 = zmm15 ^ (mem & (zmm8 ^ zmm15))
; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm7 = [6,5,8,7,0,9,0,0,6,5,8,7,0,9,0,0,6,5,8,7,0,9,0,0,6,5,8,7,0,9,0,0]
; AVX512DQ-NEXT: vpshufb %xmm7, %xmm11, %xmm9
; AVX512DQ-NEXT: vpshufb %xmm7, %xmm12, %xmm10
@@ -2863,30 +2867,32 @@ define void @store_i8_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm9, %zmm10, %zmm9
; AVX512DQ-NEXT: vpermq {{.*#+}} zmm9 = zmm9[0,0,0,1,4,4,4,5]
; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm9 = zmm9 ^ (zmm14 & (zmm9 ^ zmm8))
-; AVX512DQ-NEXT: vpshufb %ymm15, %ymm4, %ymm8
-; AVX512DQ-NEXT: vpshufb %ymm15, %ymm2, %ymm10
-; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} ymm8 = ymm10[0],ymm8[0],ymm10[1],ymm8[1],ymm10[2],ymm8[2],ymm10[3],ymm8[3],ymm10[4],ymm8[4],ymm10[5],ymm8[5],ymm10[6],ymm8[6],ymm10[7],ymm8[7],ymm10[16],ymm8[16],ymm10[17],ymm8[17],ymm10[18],ymm8[18],ymm10[19],ymm8[19],ymm10[20],ymm8[20],ymm10[21],ymm8[21],ymm10[22],ymm8[22],ymm10[23],ymm8[23]
-; AVX512DQ-NEXT: vpunpckhbw {{.*#+}} ymm10 = ymm3[8],ymm5[8],ymm3[9],ymm5[9],ymm3[10],ymm5[10],ymm3[11],ymm5[11],ymm3[12],ymm5[12],ymm3[13],ymm5[13],ymm3[14],ymm5[14],ymm3[15],ymm5[15],ymm3[24],ymm5[24],ymm3[25],ymm5[25],ymm3[26],ymm5[26],ymm3[27],ymm5[27],ymm3[28],ymm5[28],ymm3[29],ymm5[29],ymm3[30],ymm5[30],ymm3[31],ymm5[31]
-; AVX512DQ-NEXT: vpshufb {{.*#+}} ymm10 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,22,23,28,29,26,27,30,31,30,31,30,31,30,31]
+; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm8 = [8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0]
+; AVX512DQ-NEXT: vpshufb %ymm8, %ymm3, %ymm10
+; AVX512DQ-NEXT: vpshufb %ymm8, %ymm2, %ymm8
+; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} ymm8 = ymm8[0],ymm10[0],ymm8[1],ymm10[1],ymm8[2],ymm10[2],ymm8[3],ymm10[3],ymm8[4],ymm10[4],ymm8[5],ymm10[5],ymm8[6],ymm10[6],ymm8[7],ymm10[7],ymm8[16],ymm10[16],ymm8[17],ymm10[17],ymm8[18],ymm10[18],ymm8[19],ymm10[19],ymm8[20],ymm10[20],ymm8[21],ymm10[21],ymm8[22],ymm10[22],ymm8[23],ymm10[23]
+; AVX512DQ-NEXT: vpunpckhbw {{.*#+}} ymm10 = ymm4[8],ymm5[8],ymm4[9],ymm5[9],ymm4[10],ymm5[10],ymm4[11],ymm5[11],ymm4[12],ymm5[12],ymm4[13],ymm5[13],ymm4[14],ymm5[14],ymm4[15],ymm5[15],ymm4[24],ymm5[24],ymm4[25],ymm5[25],ymm4[26],ymm5[26],ymm4[27],ymm5[27],ymm4[28],ymm5[28],ymm4[29],ymm5[29],ymm4[30],ymm5[30],ymm4[31],ymm5[31]
+; AVX512DQ-NEXT: vpshufb {{.*#+}} ymm10 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,22,23,28,29,30,31,30,31,30,31,30,31]
; AVX512DQ-NEXT: vinserti64x4 $1, %ymm10, %zmm8, %zmm8
; AVX512DQ-NEXT: vpermq {{.*#+}} zmm8 = zmm8[2,2,2,3,6,6,6,7]
-; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm10 = [8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0]
-; AVX512DQ-NEXT: vpshufb %ymm10, %ymm5, %ymm5
-; AVX512DQ-NEXT: vpshufb %ymm10, %ymm3, %ymm3
-; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} ymm3 = ymm3[0],ymm5[0],ymm3[1],ymm5[1],ymm3[2],ymm5[2],ymm3[3],ymm5[3],ymm3[4],ymm5[4],ymm3[5],ymm5[5],ymm3[6],ymm5[6],ymm3[7],ymm5[7],ymm3[16],ymm5[16],ymm3[17],ymm5[17],ymm3[18],ymm5[18],ymm3[19],ymm5[19],ymm3[20],ymm5[20],ymm3[21],ymm5[21],ymm3[22],ymm5[22],ymm3[23],ymm5[23]
-; AVX512DQ-NEXT: vpunpckhbw {{.*#+}} ymm2 = ymm2[8],ymm4[8],ymm2[9],ymm4[9],ymm2[10],ymm4[10],ymm2[11],ymm4[11],ymm2[12],ymm4[12],ymm2[13],ymm4[13],ymm2[14],ymm4[14],ymm2[15],ymm4[15],ymm2[24],ymm4[24],ymm2[25],ymm4[25],ymm2[26],ymm4[26],ymm2[27],ymm4[27],ymm2[28],ymm4[28],ymm2[29],ymm4[29],ymm2[30],ymm4[30],ymm2[31],ymm4[31]
-; AVX512DQ-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,22,23,28,29,30,31,30,31,30,31,30,31]
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2
+; AVX512DQ-NEXT: vpshufb %ymm13, %ymm5, %ymm5
+; AVX512DQ-NEXT: vpshufb %ymm13, %ymm4, %ymm4
+; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} ymm4 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[4],ymm5[4],ymm4[5],ymm5[5],ymm4[6],ymm5[6],ymm4[7],ymm5[7],ymm4[16],ymm5[16],ymm4[17],ymm5[17],ymm4[18],ymm5[18],ymm4[19],ymm5[19],ymm4[20],ymm5[20],ymm4[21],ymm5[21],ymm4[22],ymm5[22],ymm4[23],ymm5[23]
+; AVX512DQ-NEXT: vpunpckhbw {{.*#+}} ymm2 = ymm2[8],ymm3[8],ymm2[9],ymm3[9],ymm2[10],ymm3[10],ymm2[11],ymm3[11],ymm2[12],ymm3[12],ymm2[13],ymm3[13],ymm2[14],ymm3[14],ymm2[15],ymm3[15],ymm2[24],ymm3[24],ymm2[25],ymm3[25],ymm2[26],ymm3[26],ymm2[27],ymm3[27],ymm2[28],ymm3[28],ymm2[29],ymm3[29],ymm2[30],ymm3[30],ymm2[31],ymm3[31]
+; AVX512DQ-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,22,23,28,29,26,27,30,31,30,31,30,31,30,31]
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm2, %zmm4, %zmm2
; AVX512DQ-NEXT: vpermq {{.*#+}} zmm2 = zmm2[2,2,2,3,6,6,6,7]
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm2 = zmm2 ^ (mem & (zmm2 ^ zmm8))
-; AVX512DQ-NEXT: vpshufb %ymm7, %ymm1, %ymm3
+; AVX512DQ-NEXT: vbroadcasti64x4 {{.*#+}} zmm3 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512DQ-NEXT: # zmm3 = mem[0,1,2,3,0,1,2,3]
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm3 = zmm8 ^ (zmm3 & (zmm2 ^ zmm8))
+; AVX512DQ-NEXT: vpshufb %ymm7, %ymm1, %ymm2
; AVX512DQ-NEXT: vpshufb %ymm7, %ymm0, %ymm4
-; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[4],ymm3[4],ymm4[5],ymm3[5],ymm4[6],ymm3[6],ymm4[7],ymm3[7],ymm4[16],ymm3[16],ymm4[17],ymm3[17],ymm4[18],ymm3[18],ymm4[19],ymm3[19],ymm4[20],ymm3[20],ymm4[21],ymm3[21],ymm4[22],ymm3[22],ymm4[23],ymm3[23]
+; AVX512DQ-NEXT: vpunpcklbw {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[4],ymm2[4],ymm4[5],ymm2[5],ymm4[6],ymm2[6],ymm4[7],ymm2[7],ymm4[16],ymm2[16],ymm4[17],ymm2[17],ymm4[18],ymm2[18],ymm4[19],ymm2[19],ymm4[20],ymm2[20],ymm4[21],ymm2[21],ymm4[22],ymm2[22],ymm4[23],ymm2[23]
; AVX512DQ-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
; AVX512DQ-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27,24,25,22,23,28,29,26,27,28,29,30,31]
-; AVX512DQ-NEXT: vinserti64x4 $1, %ymm0, %zmm3, %zmm0
+; AVX512DQ-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm0
; AVX512DQ-NEXT: vpermq {{.*#+}} zmm0 = zmm0[2,2,2,3,6,6,6,7]
-; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm2))
+; AVX512DQ-NEXT: vpternlogq {{.*#+}} zmm0 = zmm0 ^ (mem & (zmm0 ^ zmm3))
; AVX512DQ-NEXT: vmovdqa64 %zmm0, 128(%rax)
; AVX512DQ-NEXT: vmovdqa64 %zmm9, (%rax)
; AVX512DQ-NEXT: vmovdqa64 %zmm6, 64(%rax)
@@ -2902,32 +2908,34 @@ define void @store_i8_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %ymm5
; AVX512DQ-FCP-NEXT: vmovdqa (%r8), %ymm0
; AVX512DQ-FCP-NEXT: vmovdqa (%r9), %ymm1
-; AVX512DQ-FCP-NEXT: vpbroadcastq {{.*#+}} ymm7 = [5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm5, %ymm6
-; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm4, %ymm8
-; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} ymm6 = ymm8[0],ymm6[0],ymm8[1],ymm6[1],ymm8[2],ymm6[2],ymm8[3],ymm6[3],ymm8[4],ymm6[4],ymm8[5],ymm6[5],ymm8[6],ymm6[6],ymm8[7],ymm6[7],ymm8[16],ymm6[16],ymm8[17],ymm6[17],ymm8[18],ymm6[18],ymm8[19],ymm6[19],ymm8[20],ymm6[20],ymm8[21],ymm6[21],ymm8[22],ymm6[22],ymm8[23],ymm6[23]
-; AVX512DQ-FCP-NEXT: vpunpckhbw {{.*#+}} ymm8 = ymm2[8],ymm3[8],ymm2[9],ymm3[9],ymm2[10],ymm3[10],ymm2[11],ymm3[11],ymm2[12],ymm3[12],ymm2[13],ymm3[13],ymm2[14],ymm3[14],ymm2[15],ymm3[15],ymm2[24],ymm3[24],ymm2[25],ymm3[25],ymm2[26],ymm3[26],ymm2[27],ymm3[27],ymm2[28],ymm3[28],ymm2[29],ymm3[29],ymm2[30],ymm3[30],ymm2[31],ymm3[31]
-; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,22,23,28,29,26,27,30,31,30,31,30,31,30,31]
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm6, %zmm6
+; AVX512DQ-FCP-NEXT: vpbroadcastq {{.*#+}} ymm6 = [8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm3, %ymm7
+; AVX512DQ-FCP-NEXT: vpshufb %ymm6, %ymm2, %ymm6
+; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[1],ymm7[1],ymm6[2],ymm7[2],ymm6[3],ymm7[3],ymm6[4],ymm7[4],ymm6[5],ymm7[5],ymm6[6],ymm7[6],ymm6[7],ymm7[7],ymm6[16],ymm7[16],ymm6[17],ymm7[17],ymm6[18],ymm7[18],ymm6[19],ymm7[19],ymm6[20],ymm7[20],ymm6[21],ymm7[21],ymm6[22],ymm7[22],ymm6[23],ymm7[23]
+; AVX512DQ-FCP-NEXT: vpunpckhbw {{.*#+}} ymm7 = ymm4[8],ymm5[8],ymm4[9],ymm5[9],ymm4[10],ymm5[10],ymm4[11],ymm5[11],ymm4[12],ymm5[12],ymm4[13],ymm5[13],ymm4[14],ymm5[14],ymm4[15],ymm5[15],ymm4[24],ymm5[24],ymm4[25],ymm5[25],ymm4[26],ymm5[26],ymm4[27],ymm5[27],ymm4[28],ymm5[28],ymm4[29],ymm5[29],ymm4[30],ymm5[30],ymm4[31],ymm5[31]
+; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,22,23,28,29,30,31,30,31,30,31,30,31]
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm7, %zmm6, %zmm6
; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} zmm6 = zmm6[2,2,2,3,6,6,6,7]
-; AVX512DQ-FCP-NEXT: vpbroadcastq {{.*#+}} ymm8 = [8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0,8,7,6,9,0,0,10,0]
-; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm3, %ymm9
-; AVX512DQ-FCP-NEXT: vpshufb %ymm8, %ymm2, %ymm8
-; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} ymm8 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[4],ymm9[4],ymm8[5],ymm9[5],ymm8[6],ymm9[6],ymm8[7],ymm9[7],ymm8[16],ymm9[16],ymm8[17],ymm9[17],ymm8[18],ymm9[18],ymm8[19],ymm9[19],ymm8[20],ymm9[20],ymm8[21],ymm9[21],ymm8[22],ymm9[22],ymm8[23],ymm9[23]
-; AVX512DQ-FCP-NEXT: vpunpckhbw {{.*#+}} ymm9 = ymm4[8],ymm5[8],ymm4[9],ymm5[9],ymm4[10],ymm5[10],ymm4[11],ymm5[11],ymm4[12],ymm5[12],ymm4[13],ymm5[13],ymm4[14],ymm5[14],ymm4[15],ymm5[15],ymm4[24],ymm5[24],ymm4[25],ymm5[25],ymm4[26],ymm5[26],ymm4[27],ymm5[27],ymm4[28],ymm5[28],ymm4[29],ymm5[29],ymm4[30],ymm5[30],ymm4[31],ymm5[31]
-; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,22,23,28,29,30,31,30,31,30,31,30,31]
+; AVX512DQ-FCP-NEXT: vpbroadcastq {{.*#+}} ymm7 = [5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10,5,8,7,6,9,0,0,10]
+; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm5, %ymm8
+; AVX512DQ-FCP-NEXT: vpshufb %ymm7, %ymm4, %ymm9
+; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} ymm8 = ymm9[0],ymm8[0],ymm9[1],ymm8[1],ymm9[2],ymm8[2],ymm9[3],ymm8[3],ymm9[4],ymm8[4],ymm9[5],ymm8[5],ymm9[6],ymm8[6],ymm9[7],ymm8[7],ymm9[16],ymm8[16],ymm9[17],ymm8[17],ymm9[18],ymm8[18],ymm9[19],ymm8[19],ymm9[20],ymm8[20],ymm9[21],ymm8[21],ymm9[22],ymm8[22],ymm9[23],ymm8[23]
+; AVX512DQ-FCP-NEXT: vpunpckhbw {{.*#+}} ymm9 = ymm2[8],ymm3[8],ymm2[9],ymm3[9],ymm2[10],ymm3[10],ymm2[11],ymm3[11],ymm2[12],ymm3[12],ymm2[13],ymm3[13],ymm2[14],ymm3[14],ymm2[15],ymm3[15],ymm2[24],ymm3[24],ymm2[25],ymm3[25],ymm2[26],ymm3[26],ymm2[27],ymm3[27],ymm2[28],ymm3[28],ymm2[29],ymm3[29],ymm2[30],ymm3[30],ymm2[31],ymm3[31]
+; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,22,23,28,29,26,27,30,31,30,31,30,31,30,31]
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm9, %zmm8, %zmm8
; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} zmm8 = zmm8[2,2,2,3,6,6,6,7]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm8 = zmm8 ^ (mem & (zmm8 ^ zmm6))
+; AVX512DQ-FCP-NEXT: vbroadcasti64x4 {{.*#+}} zmm9 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512DQ-FCP-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3]
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm9 = zmm6 ^ (zmm9 & (zmm8 ^ zmm6))
; AVX512DQ-FCP-NEXT: vpbroadcastq {{.*#+}} ymm10 = [6,5,8,7,0,9,0,0,6,5,8,7,0,9,0,0,6,5,8,7,0,9,0,0,6,5,8,7,0,9,0,0]
; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm1, %ymm6
-; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm0, %ymm9
-; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} ymm6 = ymm9[0],ymm6[0],ymm9[1],ymm6[1],ymm9[2],ymm6[2],ymm9[3],ymm6[3],ymm9[4],ymm6[4],ymm9[5],ymm6[5],ymm9[6],ymm6[6],ymm9[7],ymm6[7],ymm9[16],ymm6[16],ymm9[17],ymm6[17],ymm9[18],ymm6[18],ymm9[19],ymm6[19],ymm9[20],ymm6[20],ymm9[21],ymm6[21],ymm9[22],ymm6[22],ymm9[23],ymm6[23]
-; AVX512DQ-FCP-NEXT: vpunpckhbw {{.*#+}} ymm9 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
-; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27,24,25,22,23,28,29,26,27,28,29,30,31]
-; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm9, %zmm6, %zmm6
+; AVX512DQ-FCP-NEXT: vpshufb %ymm10, %ymm0, %ymm8
+; AVX512DQ-FCP-NEXT: vpunpcklbw {{.*#+}} ymm6 = ymm8[0],ymm6[0],ymm8[1],ymm6[1],ymm8[2],ymm6[2],ymm8[3],ymm6[3],ymm8[4],ymm6[4],ymm8[5],ymm6[5],ymm8[6],ymm6[6],ymm8[7],ymm6[7],ymm8[16],ymm6[16],ymm8[17],ymm6[17],ymm8[18],ymm6[18],ymm8[19],ymm6[19],ymm8[20],ymm6[20],ymm8[21],ymm6[21],ymm8[22],ymm6[22],ymm8[23],ymm6[23]
+; AVX512DQ-FCP-NEXT: vpunpckhbw {{.*#+}} ymm8 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
+; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27,24,25,22,23,28,29,26,27,28,29,30,31]
+; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm8, %zmm6, %zmm6
; AVX512DQ-FCP-NEXT: vpermq {{.*#+}} zmm6 = zmm6[2,2,2,3,6,6,6,7]
-; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm6 ^ (mem & (zmm6 ^ zmm8))
+; AVX512DQ-FCP-NEXT: vpternlogq {{.*#+}} zmm6 = zmm6 ^ (mem & (zmm6 ^ zmm9))
; AVX512DQ-FCP-NEXT: vmovdqa (%rcx), %xmm9
; AVX512DQ-FCP-NEXT: vpshufb %xmm7, %xmm9, %xmm8
; AVX512DQ-FCP-NEXT: vmovdqa (%rdx), %xmm11
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
index 9825c87ee069c..9aa3330c57778 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
@@ -9034,7 +9034,7 @@ define void @store_i8_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-FCP-NEXT: vshufi64x2 {{.*#+}} zmm9 = zmm2[0,1,0,1],zmm0[0,1,0,1]
; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm0 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,22,23,26,27,24,25,22,23,24,25,26,27,26,27,24,25]
; AVX512-FCP-NEXT: vpshuflw {{.*#+}} ymm15 = ymm13[2,1,1,2,4,5,6,7,10,9,9,10,12,13,14,15]
-; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm28 = [4,5,4,5,5,7,4,5,20,21,22,23,20,21,22,23]
+; AVX512-FCP-NEXT: vpmovsxbd {{.*#+}} zmm28 = [0,5,4,0,5,0,4,0,20,21,0,23,0,21,0,23]
; AVX512-FCP-NEXT: vpermt2d %zmm0, %zmm28, %zmm15
; AVX512-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX512-FCP-NEXT: vmovdqa64 %ymm29, %ymm0
@@ -9761,7 +9761,7 @@ define void @store_i8_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-FCP-NEXT: vshufi64x2 {{.*#+}} zmm9 = zmm2[0,1,0,1],zmm0[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm0 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,22,23,26,27,24,25,22,23,24,25,26,27,26,27,24,25]
; AVX512DQ-FCP-NEXT: vpshuflw {{.*#+}} ymm15 = ymm13[2,1,1,2,4,5,6,7,10,9,9,10,12,13,14,15]
-; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm28 = [4,5,4,5,5,7,4,5,20,21,22,23,20,21,22,23]
+; AVX512DQ-FCP-NEXT: vpmovsxbd {{.*#+}} zmm28 = [0,5,4,0,5,0,4,0,20,21,0,23,0,21,0,23]
; AVX512DQ-FCP-NEXT: vpermt2d %zmm0, %zmm28, %zmm15
; AVX512DQ-FCP-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX512DQ-FCP-NEXT: vmovdqa64 %ymm29, %ymm0
diff --git a/llvm/test/CodeGen/X86/vselect-pcmp.ll b/llvm/test/CodeGen/X86/vselect-pcmp.ll
index eecb298ca3ec4..ab487ed888981 100644
--- a/llvm/test/CodeGen/X86/vselect-pcmp.ll
+++ b/llvm/test/CodeGen/X86/vselect-pcmp.ll
@@ -1558,7 +1558,7 @@ define void @store_blend_load_v16i16(ptr %a0, ptr %a1, ptr %a2) {
; XOP-NEXT: vmovdqa (%rdi), %ymm0
; XOP-NEXT: vbroadcastss {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7]
; XOP-NEXT: vpcomltuw 16(%rdi), %xmm1, %xmm2
-; XOP-NEXT: vpcomltuw (%rdi), %xmm1, %xmm1
+; XOP-NEXT: vpcomgtuw %xmm1, %xmm0, %xmm1
; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; XOP-NEXT: vpcmov %ymm1, (%rsi), %ymm0, %ymm0
; XOP-NEXT: vmovdqa %ymm0, (%rdx)
@@ -1627,7 +1627,7 @@ define void @store_blend_load_v32i8(ptr %a0, ptr %a1, ptr %a2) {
; XOP-NEXT: vmovdqa (%rdi), %ymm0
; XOP-NEXT: vbroadcastss {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
; XOP-NEXT: vpcomltub 16(%rdi), %xmm1, %xmm2
-; XOP-NEXT: vpcomltub (%rdi), %xmm1, %xmm1
+; XOP-NEXT: vpcomgtub %xmm1, %xmm0, %xmm1
; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; XOP-NEXT: vpcmov %ymm1, (%rsi), %ymm0, %ymm0
; XOP-NEXT: vmovdqa %ymm0, (%rdx)
More information about the llvm-commits
mailing list